Zbornik 25. mednarodne multikonference INFORMACIJSKA DRUZBA Zvezek B Proceedings of the 25th International Multiconference INFORMATION SOCIETY Volume B Kognitivna znanost Cognitive Science Uredniki  Editors: Toma Strle, Borut Trpin, Olga Markic Ljubljana, Slovenija 13. oktober 13 October Ljubljana, Slovenia httpis.ijs.si Zbornik 25. mednarodne multikonference INFORMACIJSKA DRUŽBA – IS 2022 Zvezek B Proceedings of the 25th International Multiconference INFORMATION SOCIETY – IS 2022 Volume B Kognitivna znanost Cognitive Science Uredniki / Editors Toma Strle, Borut Trpin, Olga Markič http://is.ijs.si 13. oktober 2022 / 13 October 2022 Ljubljana, Slovenija Uredniki: Toma Strle Center za Kognitivno znanost, Pedagoška fakulteta, Univerza v Ljubljani Borut Trpin Filozofska fakulteta, Univerza v Ljubljani Olga Markič Filozofska fakulteta, Univerza v Ljubljani Založnik: Institut »Jožef Stefan«, Ljubljana Priprava zbornika: Mitja Lasič, Vesna Lasič, Lana Zemljak Oblikovanje naslovnice: Vesna Lasič Dostop do e-publikacije: http://library.ijs.si/Stacks/Proceedings/InformationSociety Ljubljana, oktober 2022 Informacijska družba ISSN 2630-371X Kataložni zapis o publikaciji (CIP) pripravili v Narodni in univerzitetni knjižnici v Ljubljani COBISS.SI-ID 127437571 ISBN 978-961-264-242-6 (PDF) PREDGOVOR MULTIKONFERENCI INFORMACIJSKA DRUŽBA 2022 Petindvajseta multikonferenca Informacijska družba je preživela probleme zaradi korone. Zahvala za skoraj normalno delovanje konference gre predvsem tistim predsednikom konferenc, ki so kljub prvi pandemiji modernega sveta pogumno obdržali visok strokovni nivo. Pandemija v letih 2020 do danes skoraj v ničemer ni omejila neverjetne rasti IKTja, informacijske družbe, umetne inteligence in znanosti nasploh, ampak nasprotno – rast znanja, računalništva in umetne inteligence se nadaljuje z že kar običajno nesluteno hitrostjo. Po drugi strani se nadaljuje razpadanje družbenih vrednot ter tragična vojna v Ukrajini, ki lahko pljuskne v Evropo. Se pa zavedanje večine ljudi, da je potrebno podpreti stroko, krepi. Konec koncev je v 2022 v veljavo stopil not raziskovalni zakon, ki bo izboljšal razmere, predvsem leto za letom povečeval sredstva za znanost. Letos smo v multikonferenco povezali enajst odličnih neodvisnih konferenc, med njimi »Legende računalništva«, s katero postavljamo nov mehanizem promocije informacijske družbe. IS 2022 zajema okoli 200 predstavitev, povzetkov in referatov v okviru samostojnih konferenc in delavnic ter 400 obiskovalcev. Prireditev so spremljale okrogle mize in razprave ter posebni dogodki, kot je svečana podelitev nagrad. Izbrani prispevki bodo izšli tudi v posebni številki revije Informatica (http://www.informatica.si/), ki se ponaša s 46-letno tradicijo odlične znanstvene revije. Multikonferenco Informacijska družba 2022 sestavljajo naslednje samostojne konference: • Slovenska konferenca o umetni inteligenci • Izkopavanje znanja in podatkovna skladišča • Demografske in družinske analize • Kognitivna znanost • Kognitonika • Legende računalništva • Vseprisotne zdravstvene storitve in pametni senzorji • Mednarodna konferenca o prenosu tehnologij • Vzgoja in izobraževanje v informacijski družbi • Študentska konferenca o računalniškem raziskovanju • Matcos 2022 Soorganizatorji in podporniki konference so različne raziskovalne institucije in združenja, med njimi ACM Slovenija, SLAIS, DKZ in druga slovenska nacionalna akademija, Inženirska akademija Slovenije (IAS). V imenu organizatorjev konference se zahvaljujemo združenjem in institucijam, še posebej pa udeležencem za njihove dragocene prispevke in priložnost, da z nami delijo svoje izkušnje o informacijski družbi. Zahvaljujemo se tudi recenzentom za njihovo pomoč pri recenziranju. S podelitvijo nagrad, še posebej z nagrado Michie-Turing, se avtonomna stroka s področja opredeli do najbolj izstopajočih dosežkov. Nagrado Michie-Turing za izjemen življenjski prispevek k razvoju in promociji informacijske družbe je prejel prof. dr. Jadran Lenarčič. Priznanje za dosežek leta pripada ekipi NIJZ za portal zVEM. »Informacijsko limono« za najmanj primerno informacijsko potezo je prejela cenzura na socialnih omrežjih, »informacijsko jagodo« kot najboljšo potezo pa nova elektronska osebna izkaznica. Čestitke nagrajencem! Mojca Ciglarič, predsednik programskega odbora Matjaž Gams, predsednik organizacijskega odbora i FOREWORD - INFORMATION SOCIETY 2022 The 25th Information Society Multiconference (http://is.ijs.si) survived the COVID-19 problems. The multiconference survived due to the conference chairs who bravely decided to continue with their conferences despite the first pandemics in the modern era. The COVID-19 pandemic from 2020 till now did not decrease the growth of ICT, information society, artificial intelligence and science overall, quite on the contrary – the progress of computers, knowledge and artificial intelligence continued with the fascinating growth rate. However, the downfall of societal norms and progress seems to slowly but surely continue along with the tragical war in Ukraine. On the other hand, the awareness of the majority, that science and development are the only perspective for prosperous future, substantially grows. In 2020, a new law regulating Slovenian research was accepted promoting increase of funding year by year. The Multiconference is running parallel sessions with 200 presentations of scientific papers at twelve conferences, many round tables, workshops and award ceremonies, and 400 attendees. Among the conferences, “Legends of computing” introduce the “Hall of fame” concept for computer science and informatics. Selected papers will be published in the Informatica journal with its 46-years tradition of excellent research publishing. The Information Society 2022 Multiconference consists of the following conferences: • Slovenian Conference on Artificial Intelligence • Data Mining and Data Warehouses • Cognitive Science • Demographic and family analyses • Cognitonics • Legends of computing • Pervasive health and smart sensing • International technology transfer conference • Education in information society • Student computer science research conference 2022 • Matcos 2022 The multiconference is co-organized and supported by several major research institutions and societies, among them ACM Slovenia, i.e. the Slovenian chapter of the ACM, SLAIS, DKZ and the second national academy, the Slovenian Engineering Academy. In the name of the conference organizers, we thank all the societies and institutions, and particularly all the participants for their valuable contribution and their interest in this event, and the reviewers for their thorough reviews. The award for life-long outstanding contributions is presented in memory of Donald Michie and Alan Turing. The Michie-Turing award was given to Prof. Dr. Jadran Lenarčič for his life-long outstanding contribution to the development and promotion of information society in our country. In addition, the yearly recognition for current achievements was awarded to NIJZ for the zVEM platform. The information lemon goes to the censorship on social networks. The information strawberry as the best information service last year went to the electronic identity card. Congratulations! Mojca Ciglarič, Programme Committee Chair Matjaž Gams, Organizing Committee Chair ii KONFERENČNI ODBORI CONFERENCE COMMITTEES International Programme Committee Organizing Committee Vladimir Bajic, South Africa Matjaž Gams, chair Heiner Benking, Germany Mitja Luštrek Se Woo Cheon, South Korea Lana Zemljak Howie Firth, UK Vesna Koricki Olga Fomichova, Russia Mitja Lasič Vladimir Fomichov, Russia Blaž Mahnič Vesna Hljuz Dobric, Croatia Alfred Inselberg, Israel Jay Liebowitz, USA Huan Liu, Singapore Henz Martin, Germany Marcin Paprzycki, USA Claude Sammut, Australia Jiri Wiedermann, Czech Republic Xindong Wu, USA Yiming Ye, USA Ning Zhong, USA Wray Buntine, Australia Bezalel Gavish, USA Gal A. Kaminka, Israel Mike Bain, Australia Michela Milano, Italy Derong Liu, Chicago, USA Toby Walsh, Australia Sergio Campos-Cordobes, Spain Shabnam Farahmand, Finland Sergio Crovella, Italy Programme Committee Mojca Ciglarič, chair Nikola Guid Andrej Ule Bojan Orel, Marjan Heričko Boštjan Vilfan Franc Solina, Borka Jerman Blažič Džonova Baldomir Zajc Viljan Mahnič, Gorazd Kandus Blaž Zupan Cene Bavec, Urban Kordeš Boris Žemva Tomaž Kalin, Marjan Krisper Leon Žlajpah Jozsef Györkös, Andrej Kuščer Niko Zimic Tadej Bajd Jadran Lenarčič Rok Piltaver Jaroslav Berce Borut Likar Toma Strle Mojca Bernik Janez Malačič Tine Kolenik Marko Bohanec Olga Markič Franci Pivec Ivan Bratko Dunja Mladenič Uroš Rajkovič Andrej Brodnik Franc Novak Borut Batagelj Dušan Caf Vladislav Rajkovič Tomaž Ogrin Saša Divjak Grega Repovš Aleš Ude Tomaž Erjavec Ivan Rozman Bojan Blažica Bogdan Filipič Niko Schlamberger Matjaž Kljun Andrej Gams Stanko Strmčnik Robert Blatnik Matjaž Gams Jurij Šilc Erik Dovgan Mitja Luštrek Jurij Tasič Špela Stres Marko Grobelnik Denis Trček Anton Gradišek iii iv KAZALO / TABLE OF CONTENTS Kognitivna znanost / Cognitive Science .................................................................................................................. 1 PREDGOVOR / FOREWORD ................................................................................................................................. 3 PROGRAMSKI ODBORI / PROGRAMME COMMITTEES ..................................................................................... 4 Into the Constant Now—Comparing DES and micro- phenomenology, two methods for exploring consciousness / Bass-Krueger Julian, Wiedemann Elisa, Demšar Ema ................................................................................... 5 LTP and LTD dependence on spontaneous activity in hippocampal and cortical glutamate synapses and the role of anaesthetics in the study of plasticity and learning / Bratuša Maša ............................................................ 10 Trusted sources and disinformation: studying the limits of science / Gsenger Rita ............................................. 14 Opacity and understanding in artificial neural networks: a philosophical perspective / Justin Martin .................. 18 Politizirana znanost in zaupanje v znanost kot politična uniforma / Marušič Jar Žiga ......................................... 22 Filozofski in psihološki vidiki človeške racionalnosti / Tomat Nastja .................................................................... 27 Joint history of play provides means for coordination / Voronina Liubov, Heintz Christophe .............................. 33 Predicting Trust in Science in the Context of COVID-19 Pandemic: The Role of Sociodemographics and Social Media Use / Zelič Žan, Berič Martin, Kobal Grum Darja ................................................................................. 37 Indeks avtorjev / Author index ................................................................................................................................ 41 v vi Zbornik 25. mednarodne multikonference INFORMACIJSKA DRUŽBA – IS 2022 Zvezek B Proceedings of the 25th International Multiconference INFORMATION SOCIETY – IS 2022 Volume B Kognitivna znanost Cognitive Science Uredniki / Editors Toma Strle, Borut Trpin, Olga Markič http://is.ijs.si 13. oktober 2022 / 13 October 2022 Ljubljana, Slovenija 1 2 PREDGOVOR Na tokratni konferenci Kognitivna znanost sodelujejo avtorice in avtorji, ki se raziskovalno ukvarjajo s kognitivno znanostjo, in predstavljajo tako empirične rezultate svojih raziskav kot tudi teoretska raziskovanja z najrazličnejših področij – od psihologije in nevroznanosti do filozofije in umetne inteligence. Poseben poudarek na letošnji konferenci posvečamo kognitivnim vidikom zaupanja v znanost, kar avtorice in avtorji naslavljajo tako z družbenega, političnega, psihološkega in filozofskega vidika. Upamo, da bo letošnja disciplinarno in metodološko bogata konferenca odprla prostor za povezovanje pronicljivih idej ter povezala domače in tuje, mlade in izkušene znanstvenice in znanstvenike, ki se ukvarjajo z vprašanji kognicije. Borut Trpin Toma Strle Olga Markič FOREWORD At this year’s Cognitive Science conference, the authors who actively research in scope of cognitive science present their empirical studies as well as theoretical research from a diverse range of disciplinary backgrounds – from psychology and neuroscience to philosophy and artificial intelligence. A special focus of this year’s conference is on cognitive aspects of trust in science. The authors address this topic from a social, political, psychological, and philosophical viewpoint. We hope that this year's cognitive science conference – rich in disciplinary approaches and methodologies – will open space for exchanging intriguing research ideas and will bring together local and international, junior and senior scientists from a diverse range of areas related to the exploration of the human mind. Borut Trpin Toma Strle Olga Markič 3 PROGRAMSKI ODBOR / PROGRAMME COMMITTEE Toma Strle, Center za Kognitivno znanost, Pedagoška fakulteta, Univerza v Ljubljani Borut Trpin, Filozofska fakulteta, Univerza v Ljubljani Olga Markič, Filozofska fakulteta, Univerza v Ljubljani Urban Kordeš, Center za kognitivno znanost, Pedagoška fakulteta, Univerza v Ljubljani Matjaž Gams, Odsek za inteligentne sisteme, Institut »Jožef Stefan«, Ljubljana 4 Into the Constant Now—Comparing DES and micro- phenomenology, two methods for exploring consciousness Julian Bass-Krueger Elisa Wiedemann Ema Demšar MSc University of Vienna Department of Cognitive Science Centre for Consciousness and Vienna, Austria Central European University PU Co ntemplative Studies julianbassk@gmail.com Vienna, Austria Monash University wiedemann_elisa@phd.ceu.edu Melbourne, Australia em a.demsar@monash.edu ABSTRACT critiques and attempt to limit biases. Although validity cannot yet Here we compare two methods of examining conscious be proven, here we test the limits and constraints of these experience—Descriptive Experience Sampling (DES) and methods. Specifically, we look at Descriptive Experience micro-phenomenology. Both look at short episodes of Sampling, founded by Russell Hurlburt and refined with the aid experience. Both have safeguards to limit biases and distortions of fellow researchers [1]. And we’ll look at micro-from first-person reporting. But these methods are still different phenomenology—adapted by Claire Petimengin from Pierre in terms of how they deal with memory, questioning, and Vermersch’s explicitation interview [2]. analysis. In this pilot study (n=4), we use both methods in the Descriptive Experience Sampling uses random beeps to direct context of a common task. Participants were interviewed about participants towards specific, concrete episodes of experience. their experience of a mental imagery task using both methods. Micro-phenomenology guides participants to a state in which DES results focused more on fine-grained details of visual memory becomes immediate and lived. experiences. Micro-phenomenology results focused more on Both methods then use different means to aim for a common how experience extended over time, and how participants goal, of revealing short episodes of experience. Experience engaged with the task. These differences in results show that the described in the abstract is an amalgamation of warped memory, investigated methods differ in scope. To further address this, we self-perception, conceptual frames, and fleeting impressions. encourage a critical methodological pluralism where methods ‘This morning I had breakfast and felt sleepy.’ In the concrete, can continue to be improved and tested for validity. however, experience manifests as a flow of vivid nows. ‘ Now I’m watching the cream dissolve in my coffee. Now I’m picturing KEYWORDS what would happen if gravity reversed overnight and I had to Consciousness, inner experience, empirical phenomenology, rearrange my furniture on the ceiling.’ These nows, so vivid DES, micro-phenomenology when lived, can dissolve in memory like cream in coffee, so that we might forget their original color. Methods of empirical 1 BACKGROUND phenomenology aim for that color. Despite similar intentions, there has been some contention The study of first-person experience has had a difficult time. In between methods. Akhter and Hurlburt have questioned the the early 20th century a prolonged disagreement between two validity of micro-phenomenology [6]. Petitmengin has argued rival introspectionist camps led to the field’s essential banishing about DES that “the beeper is not suitable for observing very from psychology [1, 2]. A later influential study by Nisbett and brief or very fine subjective events” [7]. Is this disagreement Wilson [3] further solidified the notion that first-person data is warranted? Do methods really reveal different aspects of flawed and distorted by heuristics, overgeneralizations, and experience when used with a common task? And if so, does this memory problems. People simply don’t know what’s in their call into question the validity of one method or the other? consciousness. To give a pragmatic definition, for a conscious Methods might just have different scopes, yielding different person, there is something that it is like to be that person [4]. A results [8]. To address these questions, we compared methods conscious person might be sipping coffee noticing the rich smell with a shared task. and hearing birds chirp. An unconscious person could for example be in dreamless sleep. 2 METHODS Conscious experience takes up most of our day (presumably) and influences our identity and understanding of the world. It lies 2.1 DES behind our sensations, emotions, and thoughts. It is important. DES uses random beeps through the day to help participants And yet it’s often either assumed as trivial, approachable through better grasp their own experience. This can involve a specialized naive methods, or else unattainable, not worth even seeking to beeper or a smartphone. The participant must have an earpiece understand. New methods reject both premises–consciousness is directly in their ear throughout the procedure. The beeps are neither trivial nor unattainable. These methods attempt to delivered at randomized intervals ranging between five minutes systematize consciousness research, in a field that has been to one hour [9]. Six beeps are delivered a day. This usually takes dubbed “empirical phenomenology” [5]. They deal with past around three or four hours. In most studies, they occur during the 5 Information Society 2022, 13 October 2022, Ljubljana, Slovenia J. Bass-Krueger, E. Wiedemann, & E. Demšar participant’s daily life, not in a lab, to increase ecological elements in turn and ask questions to elicit greater specificity. validity. Micro-phenomenology aims for nuance. Questioning can often After each beep, the participant jots down notes on their inner focus on subtle emotional shifts of even shifts in body or posture experience right before the beep. So not inner experience during that contribute to experience. the beep (e.g., darn that’s annoying!) but right before. The goal There are no firm guidelines for how long a micro-is to describe that last uninterrupted moment before the beep. phenomenology interview lasts. However, it is not uncommon Usually this moment is much shorter than what participants first for short segments of experience to elicit hour-long interviews. have in mind, and can last a fraction of a second. The aim of micro-phenomenology is to uncover the complexity Questioning and training is needed in order to apprehend this and nuance of the experiential episode both at a particular moment. At the end of each day of sampling, participants are moment (synchronic dimension) and its development over time interviewed about the six beeps they collected. The interviews (diachronic dimension),with the focus of the interview last an hour and any samples not discussed within that time are depending on the research question of the particular study. discarded. There are always multiple days of sampling, usually around 5 or 6, but occasionally many more. The first day of 2.3 Main differences sampling is always discarded and considered training. Time - Micro-phenomenology typically deals with longer Subsequent days are often discarded as well, if they don’t hew to sections of time. Researchers can observe how elements change. validity criteria. Petitmengin writes, “To enter into contact with one’s experience, Validity depends primarily on participants’ ability to clearly it is necessary to respect its fluid and dynamic character” [11]. describe specific moments of experience with little hesitation and DES does also incorporate time though. Experience is not frozen equivocating language. Questioning aims to lead participants into a static snapshot. For example, if a person is innerly speaking away from generalizations. For example, a participant might first “I need to call mom” this might extend over time. And a fuzzy say, “I was driving and kinda nervous I think. I’m always nervous feeling in their chest might increase in strength over the moment. when I drive.” The use of the term ‘always’ may indicate that the The difference here is thus of degree, not of type. participant was generalizing. The use of terms ‘kinda’ and ‘I Retrospection - Micro-phenomenology, in general, involves think’ could indicate uncertainty stemming from lack of contact substantially more retrospection. The target experience could be with direct experience. Further questioning may reveal that years before the interview [11]. In DES, the target experience is experience before the beep was something completely a few seconds before the notetaking and less than 24 hours before different—perhaps a mental image of a fat squirrel with the inner the interview. There are still memory demands but they are speaking “munchy munch.” It is common in DES for results to fewer. However, as mentioned, micro-phenomenology can also go against participants’ initial expectations [9, 10]. be done with the target experience shortly before the interview 2.2 Micro-phenomenology [12]. This is the case for our comparison study. Micro-phenomenology aims to guide the participants towards Directing attention - Micro-phenomenology aims for an vividly reliving and precisely describing a past conscious episode evocation state in which participants re-live the original [7]. This episode is of underdetermined length, ranging from a experience. DES takes a more skeptical approach. DES questions few minutes to a few seconds. The episode can be in the recent encourage the participant to doubt if reported elements were past or have occurred many years ago. For the sake of bringing really part of their direct experience. DES acknowledges that this our methods as close as possible to compare them, here we’ll skepticism might lead it to miss out on elements of experience. apply micro-phenomenology to the recent past and to short But Hurlburt sees this as preferable to reporting elements that episodes (10 seconds). weren’t there [9]. Micro-phenomenology prefers having as full Memories can be indistinct, so micro-phenomenology aims to an impression of experience as possible. It offers participants guide the participant to an “evocation state” where past opportunities to revise and clarify their reports, but in service of experience is ‘re-lived’ [7]. Participants have direct contact with maintaining an evocation state, doesn’t ‘grill’ participants to the what they saw, heard, or felt at the time of the target experience. extent that DES does. Questions aim to ‘stabilize’ this evocation state and maintain the Questioning – Micro-phenomenology questioning is “non-participant’s contact with their experience. For example, inducive but directive” [7]. DES questioning is non-inducive and participants are periodically asked to return to the beginning of non-directive. For example, micro-phenomenology asks about the episode. If the participant digresses, the interviewer can specific sensory modalities in turn, i.e. ‘Do you hear anything?’ repeat the participant’s earlier descriptions. It holds that this is necessary to elicit greater detail since As in DES, participants are asked for greater specificity about participants may not know where to direct their attention. DES the elements they reveal. For example, if a participant has a would instead ask, ‘Was there anything else in your experience?’. mental image, an interviewer might ask, “Is it in colour or in In general, micro-phenomenology is more trusting of black and white? Is it detailed or fuzzy? Is it dark or light?” [7]. participant reports. DES places a greater emphasis on skepticism, Micro-phenomenology begins by eliciting the context and training participants in order to get greater fidelity. For example, sensory modalities of past experience—what participants saw, the first day of training is always discarded with DES. This is not heard, felt, etc. This helps the participant enter the evocation the case with micro-phenomenology. Training interviews are state. Once in this state, questions can be more open ended. occasionally used but optional. Interviewers can ask about the sequence of experience and how Validity - There is agreement between methods about how to different elements change over time. They then focus on specific judge validity. Both acknowledge that rules and explanations of 6 Into the Constant Now Information Society 2022, 13 October 2022, Ljubljana, Slovenia the method make their own case for validity. A successful the prompt to form mental images. They were interviewed sample/interview then depends on these guidelines being following the guidelines for micro-phenomenological interviews followed, and questions being suitably content-neutral and non- [7] after each prompt. E.W. conducted these interviews [Fig. 2]. leading. Other points of agreement include situating methods in a net of third-person observables—for example, can first-person Training data link with behavioral data? Can correlations be found with neuroimaging? No one correlation can address validity but networks of connections can help lead to first- and third- person methods informing each other through “mutual constraints” [13]. Differences include differing methods for judging veracity. Both methods rely on both verbal and non-verbal cues. But DES leans more heavily on verbal cues, like subjunctification [9]. Is the participant saying umm, I think, kindof, maybe, sorta, I Task guess? Then it’s likely they’re not describing direct experience. Micro-phenomenology relies more on visual cues—for example a participant’s eyes pointing upwards indicating that they’re in an evocation state. Petitmengin also advocates checking a participant’s reported experience against the researcher’s own experience, calling this the “kingpin of all validation” [7]. Is it similar or at least plausible? Hurlburt and Akhter [6] see this as harmful—a participants’ experience may be radically different from the Figure 1: DES training and task researcher’s, and so should be ‘bracketed’ as much as possible. 3 PROCEDURE This study involved four participants—a small sample size aimed at highlighting certain method contours rather than generalizing or making claims of statistical significance. All four were female students residing in Slovenia, aged 23 to 26. They are referred to here using pseudonyms. Each participant underwent both the DES and micro-phenomenology procedure. However, two started with micro-phenomenology and two started with DES (to limit biasing). There was a break (at least six days) before Figure 2: Micro-phenomenology training and task switching methods. To facilitate comparison, the interviews concerned participants’ experience of a task. We used a mental imagery 4 RESULTS elicitation task, in which participants were given descriptive 4.1 Similarities prompts and 10 seconds to form mental images. Examples of prompts include: “A child holds an ice cream cone with three Image characteristics - Both methods uncovered common visual scoops. The ice cream falls onto the hot pavement.” “A storm phenomena. One example of this is with GIF-like repetition. This cloud gathers over a city. A lightning bolt strikes.” may be something specific to our current digital age. These short, Before the task came training. For DES, this involved three repeating moving images are common on social media. Many days of DES sampling during the participant’s everyday life— older people in DES sampling have mental images in black and going to class, cafés, etc. Participants received six beeps a day, white [14]. The technologies of our age may shape our perception jotted down their consciousness experience in the moment before and perceptual cognition. the beep, and received hour-long interviews within 24 hours of Other commonalities include elements changing over time. sample collection [Fig. 1]. For both methods, images didn’t always emerge fully formed. For micro-phenomenology, training was much shorter. And micro-phenomenology further shows how images morphed Participants were given a task shortly before the main task—to or how new elements entered. For both methods, images could spell the word octopus. Participants were then interviewed to either be moving or static. give them some practice and familiarity with micro- With both methods, mental images had differing levels of phenomenology and the interview procedure [Fig. 2]. detail—inter- and intrasubject. Images were sometimes clear. For the task, the DES portion involved 32 pre-recorded Sometimes they were fuzzy, indistinct, ghostly, or blurry. Visual prompts. 10 seconds followed each prompt, allowing for mental elements were sometimes realistic and sometimes cartoonish. imagery formation. Five random beeps were interspersed Interactions with other modalities - Inner images could interact throughout the task, ranging from 1-10 seconds after the prompt with other types of experience. Both methods revealed words and concluded. There was a DES interview after each beep. J.B.-K. images interacting. Micro-phenomenology revealed participants conducted these interviews [Fig. 1]. sometimes innerly repeating words from the prompt. In one case, The micro-phenomenology task involved 2 prompts. These these words were a distraction from forming images. In another were on separate days. Participants again had 10 seconds after case, they spurred on a new visual perspective. 7 Information Society 2022, 13 October 2022, Ljubljana, Slovenia J. Bass-Krueger, E. Wiedemann, & E. Demšar For DES one participant misheard the word ‘chirp’ as ‘gerb’. —The task could involve constrained freedom or constraint. At the moment of the beep, she was innerly repeating it, Jelka felt constrained at times. She had to imagine things she wondering what it meant, and had a visual impression without wasn’t interested in. Anna, especially, felt freedom. She could any visual elements explicitly present—just a large mass. imagine whatever she wanted. Anna also played with the Images could also interact with feelings. DES found that 1/5 of prompts. For example, given a prompt about a boy with three samples involved feelings. These were sometimes positive in scoops of ice-cream, Anna imagined three ice-cream scoop tools. valence (‘calm’) or sometimes negative (‘dislike’). Certain We can see how she engaged with the task, lightheartedly testing prompts correlated with negative feelings—like the prompt “A how far she could push the prompts. DES could not have family gathers around the dinner table. The father starts serving revealed this entire sequence of trying out different visual food.” components. Micro-phenomenology also found feelings. For example, for a prompt about two children skating on a pond, Jelka added a 5 DISCUSSION mother to the scene and projected her own worry onto the Despite similarities, these methods have different scopes and mother. reveal different results. Micro-phenomenology revealed more temporal dynamics. We saw how images evolved over time, and 4.2 Differences how participants interacted with the prompts. DES revealed more Visual differences - While there were similarities concerning visual characteristics of images. This is contrary to Petitmengin’s mental imagery formation, there were differences as well. With comment concerning DES’s limited experiential detail: “I doubt DES, for Jelka, all 5 prompt samples involved imagery with a whether the beep enables the interviewee to direct his attention dual vantage point. She was both looking at the image from a from ‘what’ to ‘how’, unless by chance” [7]. It also goes against distance but at the same time had another vantage point of being claims from Froese, Seth, and Gould that DES adheres only to a surrounded by the scene. Think of simultaneously watching a “shallow conception of consciousness” [15]. movie on a screen and being in the movie as the main character. Note that methods differed in the treatment of fine-grained Since this dual vantage point was found in all of her samples, one details. DES revealed dual aspect imagery and micro-might expect it to be a generalizable feature of her mental phenomenology did not. This could have been the result of images. But micro-phenomenology didn’t find it. It found differing experience or a product of the research design where instances of 3rd and 1st person inner images for Jelka, but never training with one method alters reporting with the other method.1 both at the same time. Perhaps the dual vantage point was present It could also be a result of one or another method hewing more but not apprehended. closely to experience. If this is the case, we need to make sure DES findings focused more on characteristics of mental our methods are faithful. Methods that distort experience may images. lead to disagreements and stall progression of the study of —Images can have borders, no borders, or focus can be on the consciousness. For this reason, issues with retrospection, center so the participant is unsure of whether or not the image memory distortion, presuppositions, and biases need to be has edges. handled carefully. Practitioners of any method need to question —Images can be in a separate mental space or positioned over what its intent is, whether its guidelines are coherent, and what the real world, for instance on a “3D screen.” research questions it can and can’t answer. —Mental images can involve aspects that would be impossible Horizons are open for refinement of methods and in real physical space. experimentation. Emerging research is even combining elements —Two simultaneous visual spaces can be present at the same from micro-phenomenology and DES [16-19]. Oblak, for time. For example, Anna had one visual space of children skating example, combined influences from both methods for interviews on a frozen pond, and a separate space where she was creating a investigating experience during a visual-spatial memory task face to add to the children. [16]. Springinsfeld conducted micro-phenomenology inspired interviews shortly after targeted experience–aiming for Time – Micro-phenomenology focused more on experience interviews on the same day as a bulimic individual’s vomiting evolving over time. We can see how imagery changes. We can episodes, to minimize retrospection demands [17]. Caporusso see how participants interact with prompts, referring back to used DES-style beeps with an interview method hewing more them, and playing with them. We see the broader experience of closely to micro-phenomenology in order to better understand the task. sense of self and boundaries in daily life and compare this to —Some elements came naturally, others required concentration. experiences of boundary dissolution [18]. And Bass-Krueger —Elements could be disproportionate and not fit with the scene. adapted DES to a slightly wider temporal scope to investigate For example, Jelka imagined a tree with birds that didn’t fit with what is really meant by a ‘moment’ of experience [19]. Critical the rest of the scene. It was too big, and a different color. We see methodological pluralism is important going forward. We must how new elements enter and how they relate to previous acknowledge differing avenues of exploring lived experience, elements. while questioning where exactly these avenues lead us. 1 Procedurally, there seemed to have been an effect of experience with one method participant who started with the micro-phenomenological interviews and then on participants’ approach to the other (new) method of investigating experience. moved on to DES at first found the latter method ‘too skeptical’ and both required For instance, participants who started with DES and then moved on to the micro-at least as much training as participants with no prior experience with first-person phenomenological interviews approached their experience with more skepticism reporting. However, with such a small sample, it is hard to disentangle how and caution than those who started with micro-phenomenology. Conversely, one experience with one method or the other may have influenced our final results. 8 Into the Constant Now Information Society 2022, 13 October 2022, Ljubljana, Slovenia ACKNOWLEDGEMENTS [9] Russell Hurlburt and Christopher Heavey, 2006. Exploring Inner Experience: The Descriptive Experience Sampling Method. John Special thanks to Urban Kordeš and Toma Strle for the Benjamins. https://doi.org/10.1075/aicr.64 resources and guidance to carry out this project. And to our [10] Russell Hurlburt and Christopher Heavey, 2015. Investigating pristine inner experience: Implications for experience sampling and participants for the trust granted in our research. questionnaires. Consciousness and Cognition 31, 148–159. http://doi.org/10.1016/j.concog.2014.11.002 [11] Claire Petitmengin, 2011. Describing the experience of describing? The REFERENCES blindspot of introspection. Journal of Consciousness Studies 18, 1, 44-62. [12] Natalie Depraz, Maria Gyemant, and Thomas Desmidt, 2017. A first- [1] Russell Hurlburt, 1993. Sampling Inner Experience in Disturbed Affect. person analysis using third-person data as a generative method: A case Plenum Press. study of surprise in depression. Constructivist Foundations 12, 2, 191-203. [2] Pierre Vermersch, 1999. Introspection as practice. Journal of [13] Francisco Varela, 1996. Neurophenomenology: A Methodological Consciousness Studies 6, 2–3, 17–42. Remedy for the Hard Problem. Journal of Consciousness Studies 3, 4, [3] Richard E. Nisbett and Timothy D. Wilson, 1977. Telling more than we 330-49. can know: Verbal reports on mental processes. Psychological Review 84 3, [14] Russell Hurlburt, 2011. Investigating Pristine Inner Experience: Moments 231–259. https://doi.org/10.1037/0033-295x.84.3.231 of Truth. Cambridge University Press. [4] Thomas Nagel, 1974. What is it like to be a bat? The Philosophical [15] Tom Froese, Anil Seth, and Cassandra Gould, 2011. Validating and Review 84, 4, 435-450. calibrating first- and second-person methods in the science of [5] Urban Kordeš, 2016. Going beyond theory: Constructivism and empirical consciousness. Journal of Consciousness Studies 18, 2, 38-64. phenomenology. Constructivist Foundations 11, 2, 375–385. [16] Aleš Oblak, Anka Slana Ozimič, Grega Repovš, and Urban Kordeš, 2022. [6] Russell Hurlburt and Sarah Akhter, 2006. The Descriptive Experience What Individuals Experience During Visuo-Spatial Working Memory Sampling method. Phenomenology and the Cognitive Sciences 5, 271-301. Task Performance: An Exploratory Phenomenological Study. Frontiers in https://doi.org/10.1007/s11097-006-9024-0 Psychology 13. https://doi.org/10.3389/fpsyg.2022.811712 [7] Claire Petitmengin, 2006. Describing one's subjective experience in the [17] Constanze Springinsfeld [In Review]. Attention, dissociation, and feelings second person: An interview method for the science of consciousness. in bulimia nervosa [prospective title]. Mei:CogSci Master’s Thesis Phenomenology and the Cognitive Sciences 5, 229-269. [18] Jaya Caporusso, 2022. Dissolution experiences and the experience of the https://doi.org/10.1007/s11097-006-9022-2 self: An empirical phenomenological investigation. Mei:CogSci Master’s Thesis. 10.25365/thesis.71694 [8] Urban Kordeš and Ema Demšar, 2021. Horizons of becoming aware: [19] Julian Bass-Krueger, 2021. Consciousness and Time. Mei:CogSci Constructing a pragmatic-epistemological framework for empirical first-Master’s Thesis 10.25365/thesis.70232 person research. Phenomenology and the Cognitive Sciences, 1-29. https://doi.org/10.1007/s11097-021-09767-6 9 LTP and LTD dependence on spontaneous activity in hippocampal and cortical glutamate synapses and the role of anaesthetics in the study of plasticity and learning Maša Bratuša† MEi:CogSci University of Ljubljana Ljubljana Slovenia bratusa.masa@gmail.com ABSTRACT dendritic spines [2] and stimulate new AMPAr integration [3], with both of these processes being key criteria for The following article is a condensed version of a review successful LTP. paper which was motivated by the hypothesis put forward In their seminal work, Abraham and Goddard [4] showed by Benuskova and her colleagues that an ongoing pre- that there is otherwise a notable difference between and postsynaptic spontaneous activity (SA) determines homosynaptic and heterosynaptic plasticity: not only the degree of input-specific LTP elicited by “Homosynaptic plasticity occurs at synapses that were various plasticity-inducing protocols, but also the degree active during the induction. It is also called input-specific of associated LTD in neighbouring non-tetanized inputs. It or associative, governed by Hebbian-type learning rules. appears that understanding regularities of spontaneous Heterosynaptic plasticity can be induced by episodes of activity can help us define boundary conditions for both strong postsynaptic activity also at synapses that were not LTP/LTD induction and maintenance. We look into LTP active during the induction, thus making any synapse at a and LTD induction in excitatory glutamate synapses, their cell a target to heterosynaptic changes. Both forms can interrelatedness and connected non-glutamate plasticity. be induced by typical protocols and operate on the same We then assess the role of SA in plasticity and consider time scales but have differential computational properties what it means for in vitro studies where SA is limited. We and play different roles in learning systems. inquire how anaesthetics affect the general capacity for Homosynaptic plasticity mediates associative LTP and LTD induction and maintenance and we join this modifications of synaptic weights. Heterosynaptic with results on their effects on SA. All of this is taken plasticity counteracts runaway dynamics introduced by together in order to suggest protocols of notable Hebbian-type rules and balances synaptic changes." [5]. ecological validity and to provide an argument in favour of A conceptual shift in our understanding of “activity procedure standardization in the field. dependence” in heterosynaptic plasticity occured after the following experiment: Prior to stimulation the medial perforant pathway (MPP) and the lateral perforant path KEYWORDS (LPP) were equally weighted. With low-frequency Hippocampus, Cerebral Cortex, Anaesthesia, Sleep, stimulation spontaneous input activity was largely Spontaneous Activity, Synaptic Long-Term Potentiation correlated and only simultaneous or closely successive (LTP), Synaptic Long-Term Depression (LTD) spikes at these two inputs could fire the postsynaptic granule cell. Meanwhile high-frequency stimulation of the MPP decorrelated the activity between LPP and MPP, INTRODUCTION which lead to lower postsynaptic activity. Notably, there In the process of learning, there is both an increase of was no heterosynaptic LTD when the presynaptic electrochemical signalling in some synapses and a spontaneous activity was blocked [6]. This became known decrease thereof in others. Potentiation and depression as the Benuskova-Abraham model which explains include many physiological changes and are therefore "heterosynaptic" LTD as a homosynaptic phenomenon more stable over time in comparison to facilitation and due to presynaptic activity. inhibition [1]. The general understanding of Meanwhile, the baseline difference between LTP- and NMDA-dependent LTP is as follows: presynaptic LTD-inducing protocols can most simply be illustrated with stimulation opens postsynaptic NMDA channels which a difference in stimulation protocols: " 900 pulses of stimuli cause a rise in postsynaptic Ca2+. Strong depolarizations induced LTD when applied at lower frequencies (1–3 Hz), displace magnesium ions, which open more NMDA and induced LTP when applied at a higher frequency (30 channels in a positive feedback loop manner. The Hz).“ [7] postsynaptic neuron accepts even more Ca2+ ions, and All of the aforementioned considerations led researchers this superfluous concentration of Ca2+ then activates [8, 9] to investigate the role of background SA in memory CAMKII, increases cAMP and PKAII concentrations. formation. It should be noted that any activity which is not Activated CAMKII is known to increase the volume of the 10 Information Society 2022, 10–14 October 2022, Ljubljana, M. Bratuša Slovenia evoked by immediate sensory processing can be other, non-glutamate synapses – since these signalling considered as spontaneous [10, 11, 12]. The goal of our chains are extensively interdependent. Along with the review was to integrate evaluations of all known previously mentioned norepinephrine and dopamine processes that affect the animals ability to “create a receptors, endocannabinoid, GABA and various memory trace”, whether it is the physiological condition of acetylcholine receptors should be accounted for in order the animal or how the inquiry into physiological change is for us to be able to interpret and generalize our findings performed. [25]. Surveillance of tyrosine [25], serine [26], adenosine/ATP [27] and Ca2+ secretion [28] whether it be METHODS from neighbouring neurons or glial cells also appears to Data was collected from 232 peer-reviewed studies on play a vital role in outcomes of synaptic plasticity. excitatory glutamate synapses of granule cells in the Especially in the case of astrocytes, close monitoring of dentate gyrus, CA1 neurons of the hippocampus (HPC), glutamate secretion should not be neglected. As far as and cortical (CTX) networks, including those that dealt the author is aware, all of these recordings are not with developmental, pathophysiological and behavioural possible simultaneously - so a full analysis would require data. We also included computational studies of synaptic iterations of the same paradigm with different plasticity. In the process of integration, various types of permutations of controlled variables. Although genetic methodological differences had to be kept in mind. similarity of laboratory animals is regular practice, we have found evidence that conditions regarding nutrition, RESULTS and DISCUSSION activity, sleep and stress should be matched as closely as possible, as they all play a role in establishing baseline At the onset of writing we wanted to achieve a sound, stress levels and ionic/aminoacid signalling [29, 30]. precise and conclusive multivariate analysis. Yet this Stress/norepinephrine [31] minimization through ensuring numerical approach proved to be impossible due to environments that best resemble the ecological niche and overarching disparities in experimental protocols. The allow for natural behaviours is crucial both in terms of differences in methods and materials make these deriving inferences on physiological plasticity in humans experiments dissimilar to the point of barely studying the and ethical concerns. Due to dendrite [32] and button [33] same phenomenon at all, not to mention the consideration restructuring that occurs in synapses after the process of that plasticity phenomena are not a uniform class to begin learning, it would be advised to scan for their baseline with [13]. In the following sections, we are nevertheless structure since an intricate confluence of signalling chains able to provide some conclusions about which variables appears to take place at that scale. ought to be controlled for so that the experimental work is Thirdly, we have taken a stance that if we are to study ecologically valid while also giving results that are memory itself, we should focus on studies where it is available for inferences on subsequent, more complex represented as a "fully learned association with practical paradigms within the study of memory. effects" which can be doubtlessly confirmed only with Firstly, the evidence that SA plays a key role in induction experiments within behavioural paradigms [34, 35]. This and maintenance of proper strength of LTP and functionalist approach requires multiple-synapse learning concurring, homeostatic LTD is overwhelming [14, 15, with behavioural timescales (seconds-to-minutes). Not 16]. In order to provide a realistic picture of synaptic only that, but it is also unquestionably dependent on plasticity (in which SA is as natural as possible), replay during sleep [36], which means that an experiments on intact tissues should be given preference understanding of phosphorylations [37] and gene [17], since all nerve ablation limits physiological SA input. expression [38] is an indispensable part of the puzzle. If For example, when studying the CA1 region, its we are to understand memory, we ought to control for connections to CA3 [18], the dentate gyrus [19], the post-learning sleep duration and composition, but also for entorhinal CTX [20] and the medial prefrontal CTX [21] the quantity of operative gap junction [39] channels that ought to be maintained. Considering norepinephrine [7] extensively contribute to the plasticity-related signalling in and dopamine [22] modifications on glutamate-synapse sleep, both through slow oscillations and sharp-wave plasticity, there is good reason to believe that both the ripples [40]. amygdala and nucleus accumbens should remain In short, there is overwhelming evidence that SA within or connected to the HPC area under study. But when it outside the region of interest is crucial to synaptic comes to the CTX, the scope of kept projections largely plasticity in a myriad of forms (post-tetanic spiking [41], depends on the cortical region in question. Unsurprisingly, bursting [42], theta oscillations [43], slow oscillations [44] and in accordance with many authors referenced in the and sharp wave ripples [45, 46]) and that all of them full paper, a preference for in vivo recordings is advised should be taken into consideration. The more complex the [17, 23, 24]. Nevertheless, many authors agree that type of learning (declarative vs. nondeclarative, thoughtful attention to in vitro conditions could still prove behavioural sequences vs. single behaviours, simple fruitful. classical conditioning vs. nonassociative learning), the Secondly, no matter the nature of the preparation, we larger the region of interest and the more notable the would do best to also keep track of what is happening on effect of these sleep phenomena. This compounding of 11 LTP and LTD dependence on spontaneous activity in glutamate synapses of the HPC and CTX and the role of Information Society 2022, 10–14 October 2022, Ljubljana, anaesthetics Slovenia complexity also applies to most previously mentioned Aplysia.” Journal of Neuroscience 30 (16): 5781–91. signalling, as the area of messenger perfusion also https://doi.org/10.1523/jneurosci.4947-09.2010. [2] Pi, Hyun Jae, Nikolai Otmakhov, Farida El Gaamouch, David grows. Lemelin, Paul De Koninck, and John Lisman. 2010. “CaMKII Lastly, both in vivo conditions and tissue extraction Control of Spine Size and Synaptic Strength: Role of Phosphorylation States and Nonenzymatic Action.” Proceedings of demand the use of anaesthetics. Due to its equal effect the National Academy of Sciences 107 (32): 14437–42. on inhibitory and excitatory receptors, which results in https://doi.org/10.1073/pnas.1009268107. [3] Lu, W., K. Isozaki, K. W. Roche, and R. A. Nicoll. 2010. “Synaptic successful plasticity induction while also providing Targeting of AMPA Receptors Is Regulated by a CaMKII Site in sufficient insentience, application of urethane seems to be the First Intracellular Loop of GluA1.” Proceedings of the National the best option for studying plasticity, at least in adult Academy of Sciences 107 (51): 22266–71. https://doi.org/10.1073/pnas.1016289107. subjects [47]. According to previous research, isoflurane [4] Abraham, W. C., and G. V. Goddard. 1983. “Asymmetric appears to be the second best choice. There is some Relationships between Homosynaptic Long-Term Potentiation and evidence to believe that sevoflurane is a good option for Heterosynaptic Long-Term Depression.” Nature 305 (5936): 717–19. https://doi.org/10.1038/305717a0 experiments in the neonatal period [48]. There might be [5] Chistiakova, Marina, Nicholas M. Bannon, Maxim Bazhenov, and some alternatives to anaesthetic predicaments, e.g. Maxim Volgushev. 2014. “Heterosynaptic Plasticity: Multiple severing some sensory projections, usage of Mechanisms and Multiple Roles.” The Neuroscientist : A Review neurotransmitter perfusions that would correct for their Journal Bringing Neurobiology, Neurology and Psychiatry 20 (5): 483–98. https://doi.org/10.1177/1073858414529829. effects, such as norepinephrine [36], or a combination of [6] Benuskova, Lubica, and Wickliffe C. Abraham. 2007. “STDP Rule both measures. Nevertheless, a routine use of these Endowed with the BCM Sliding Threshold Accounts for remains in the realm of the hypothetical since the Hippocampal Heterosynaptic Plasticity.” Journal of Computational Neuroscience 22 (2): 129–33. bioethical committees might remain sceptical about what https://doi.org/10.1007/s10827-006-0002-x. lowering the anaesthetic dose would mean in terms of [7] Katsuki, Hiroshi, Yukitoshi Izumi, and Charles F. Zorumski. 1997. sentience and anguish [49, 50]. “Noradrenergic Regulation of Synaptic Plasticity in the At this point in time, we are far from being in possession Hippocampal CA1 Region.” Journal of Neurophysiology 77 (6): 3013–20. https://doi.org/10.1152/jn.1997.77.6.3013 of any sort of statistical law that could be considered ecologically valid even in simpler types of [8] Beňušková, Ľubica, and Peter Jedlička. 2012. “COMPUTATIONAL MODELING of LONG-TERM DEPRESSION of SYNAPTIC learning/plasticity. The author is aware that the variables WEIGHTS: INSIGHTS from STDP, METAPLASTICITY and listed in the previous sections taken together are SPONTANEOUS ACTIVITY.” Neural Network World 22 (2): 161–80. https://doi.org/10.14311/nnw.2012.22.010. essentially calling for an "ideal experiment" which is entirely unattainable within one laboratory. Yet it appears [9] Shirrafiardekani, Azam, Lubica Benuskova, and Jörg Frauendiener. 2019. “A Voltage-Based Metaplasticity Rule Applied that a combined effort of multiple institutions could to the Model Hippocampal Granule Cell Accounts for Homeostatic overcome these constraints of time and funding and make Heterosynaptic Plasticity,” February. https://doi.org/10.1101/557173. greater strides in the integration of experimental results into a cohesive body of knowledge. A collaborative search [10] Vidaurre, Diego, Laurence T. Hunt, Andrew J. Quinn, Benjamin A. E. Hunt, Matthew J. Brookes, Anna C. Nobre, and Mark W. for a law that could easily generalize across experimental Woolrich. 2018. “Spontaneous Cortical Activity Transiently conditions should most likely start with a standardisation Organises into Frequency Specific Phase-Coupling Networks.” Nature Communications 9 (1). of materials and methods and careful coordination of https://doi.org/10.1038/s41467-018-05316-z. experimental tasks within the in vitro domain of plasticity [11] Sun, W., and Y. Dan. 2009. “Layer-Specific Network Oscillation in order to gradually build up towards the end goal of and Spatiotemporal Receptive Field in the Visual Cortex.” understanding declarative memory formation. Proceedings of the National Academy of Sciences 106 (42): In conclusion, our research could only show that the 17986–91. https://doi.org/10.1073/pnas.0903962106. spectrum of phenomena contributing to various levels of [12] Florin, Esther, and Sylvain Baillet. 2015. “The Brain’s Resting-State Activity Is Shaped by Synchronized plasticity is strikingly wide and heavily interconnected - to Cross-Frequency Coupling of Neural Oscillations.” NeuroImage the point that a comprehensive understanding of learning 111 (May): 26–35. is apparently not achievable through inherently https://doi.org/10.1016/j.neuroimage.2015.01.054. untransferable results of nonpartisan research. [13] Rc, Malenka, and Bear Mf. 2004. “LTP and LTD: An Embarrassment of Riches.” Neuron. September 30, 2004. https://pubmed.ncbi.nlm.nih.gov/15450156/. ACKNOWLEDGMENTS [14] Griffiths, Benjamin J., George Parish, Frederic Roux, Sebastian The author would like to thank prof. RNDr. Ľubica Michelmann, Mircea van der Plas, Luca D. Kolibius, Ramesh Beňušková, PhD. from Department of Applied Informatics, Chelvarajah, et al. 2019. “Directional Coupling of Slow and Fast Hippocampal Gamma with Neocortical Alpha/Beta Oscillations in Faculty of Mathematics, Physics and Informatics, Human Episodic Memory.” Proceedings of the National Academy Comenius University, Slovakia for her indispensable of Sciences 116 (43): 21834–42. https://doi.org/10.1073/pnas.1914180116 mentorship in the writing of the original paper [51]. [15] Umbach, Gray, Pranish Kantak, Joshua Jacobs, Michael Kahana, REFERENCES Brad E. Pfeiffer, Michael Sperling, and Bradley Lega. 2020. “Time Cells in the Human Hippocampus and Entorhinal Cortex Support [1] Antonov, I., E. R. Kandel, and R. D. Hawkins. 2010. “Presynaptic Episodic Memory.” Proceedings of the National Academy of and Postsynaptic Mechanisms of Synaptic Plasticity and Sciences 117 (45): 28463–74. Metaplasticity during Intermediate-Term Memory Formation in https://doi.org/10.1073/pnas.2013250117. 12 Information Society 2022, 10–14 October 2022, Ljubljana, M. Bratuša Slovenia [16] Dehaene, Stanislas, and Jean-Pierre Changeux. 2005. “Ongoing Spontaneous Transmitter Release.” Neuron 29 (1): 197–208. Spontaneous Activity Controls Access to Consciousness: A https://doi.org/10.1016/s0896-6273(01)00190-8. Neuronal Model for Inattentional Blindness.” Edited by Larry [34] Cone, Ian, and Harel Z. Shouval. 2021. “Behavioral Time Scale Abbott. PLoS Biology 3 (5): e141. Plasticity of Place Fields: Mathematical Analysis.” Frontiers in https://doi.org/10.1371/journal.pbio.0030141. Computational Neuroscience 15 (March). https://doi.org/10.3389/fncom.2021.640235. [17] Antkowiak, B. 2002. “In Vitro Networks: Cortical Mechanisms of [35] Cao, Jun, Nanhui Chen, Tianle Xu, and Lin Xu. 2004. Anaesthetic Action.” British Journal of Anaesthesia 89 (1): 102–11. “Stress-Facilitated LTD Induces Output Plasticity through https://doi.org/10.1093/bja/aef154. Synchronized-Spikes and Spontaneous Unitary Discharges in the CA1 Region of the Hippocampus.” Neuroscience Research 49 (2): [18] Lin, Xiaoxiao, Michelle Amalraj, Crisylle Blanton, Brenda Avila, 229–39. https://doi.org/10.1016/j.neures.2004.03.001. Todd C. Holmes, Douglas A. Nitz, and Xiangmin Xu. 2021. [36] Reasor, Jonathan D., and Gina R. Poe. 2008. “Learning and “Noncanonical Projections to the Hippocampal CA3 Regulate Memory during Sleep and Anesthesia.” International Spatial Learning and Memory by Modulating the Feedforward Anesthesiology Clinics 46 (3): 105–29. Hippocampal Trisynaptic Pathway.” Edited by Thomas https://doi.org/10.1097/aia.0b013e318181e513. Klausberger. PLOS Biology 19 (12): e3001127. [37] Brüning, Franziska, Sara B. Noya, Tanja Bange, Stella Koutsouli, https://doi.org/10.1371/journal.pbio.3001127. Jan D. Rudolph, Shiva K. Tyagarajan, Jürgen Cox, Matthias Mann, [19] Isomura, Yoshikazu, Anton Sirota, Simal Özen, Sean Montgomery, Steven A. Brown, and Maria S. Robles. 2019. “Sleep-Wake Cycles Kenji Mizuseki, Darrell A. Henze, and György Buzsáki. 2006. Drive Daily Dynamics of Synaptic Phosphorylation.” Science 366 “Integration and Segregation of Activity in Entorhinal-Hippocampal (6462). https://doi.org/10.1126/science.aav3617. Subregions by Neocortical Slow Oscillations.” Neuron 52 (5): [38] PEKNY, T., D. ANDERSSON, U. WILHELMSSON, M. PEKNA, 871–82. https://doi.org/10.1016/j.neuron.2006.10.023. and M. PEKNY. 2014. “Short General Anaesthesia Induces [20] Hahn, Thomas T G, James M McFarland, Sven Berberich, Bert Prolonged Changes in Gene Expression in the Mouse Sakmann, and Mayank R Mehta. 2012. “Spontaneous Persistent Hippocampus.” Acta Anaesthesiologica Scandinavica 58 (9): Activity in Entorhinal Cortex Modulates Cortico-Hippocampal 1127–33. https://doi.org/10.1111/aas.12369. Interaction in Vivo.” Nature Neuroscience 15 (11): 1531–38. [39] Ross, F.M, P Gwyn, D Spanswick, and S.N Davies. 2000. https://doi.org/10.1038/nn.3236. “Carbenoxolone Depresses Spontaneous Epileptiform Activity in [21] Lundqvist, Mikael, Jonas Rose, Pawel Herman, Scott L Brincat, the CA1 Region of Rat Hippocampal Slices.” Neuroscience 100 Timothy J Buschman, and Earl K Miller. 2016. “Gamma and Beta (4): 789–96.https://doi.org/10.1016/s0306-4522(00)00346-8. Bursts Underlie Working Memory.” Neuron 90 (1): 152–64. [40] Yuste, Rafael, Alejandro Peinado, and Lawrence C. Katz. 1992. https://doi.org/10.1016/j.neuron.2016.02.028. “Neuronal Domains in Developing Neocortex.” Science 257 [22] Atherton, Laura A., David Dupret, and Jack R. Mellor. 2015. (5070): 665–69. https://doi.org/10.1126/science.1496379. “Memory Trace Replay: The Shaping of Memory Consolidation by [41] Vandael, David, Yuji Okamoto, and Peter Jonas. 2021. Neuromodulation.” Trends in Neurosciences 38 (9): 560–70. “Transsynaptic Modulation of Presynaptic Short-Term Plasticity in https://doi.org/10.1016/j.tins.2015.07.004. Hippocampal Mossy Fiber Synapses.” Nature Communications 12 [23] Whittington, Miles A., Mark O. Cunningham, Fiona E.N. LeBeau, (1). https://doi.org/10.1038/s41467-021-23153-5. Claudia Racca, and Roger D. Traub. 2010. “Multiple Origins of the [42] Goodman, Corey S., and Carla J. Shatz. 1993. “Developmental Cortical Gamma Rhythm.” Developmental Neurobiology 71 (1): Mechanisms That Generate Precise Patterns of Neuronal 92–106. https://doi.org/10.1002/dneu.20814. Connectivity.” Cell 72 (January): 77–98. [24] Pollard, Christopher E., and Anthony Angel. 1990. “Spontaneous https://doi.org/10.1016/s0092-8674(05)80030-3. Single Cell Discharge in Rat Somatosensory Cortical Slices and [43] Wang, Mengni, David J. Foster, and Brad E. Pfeiffer. 2020. Its Relationship to Discharge in the Urethane-Anaesthetized Rat.” “Alternating Sequences of Future and Past Behavior Encoded Brain Research 518 (1-2): 120–26. within Hippocampal Theta Oscillations.” Science 370 (6513): https://doi.org/10.1016/0006-8993(90)90962-b. 247–50. https://doi.org/10.1126/science.abb4151. [25] Bashir, Zafar I. 2003. “On Long-Term Depression Induced by [44] Kim, Jaekyung, Tanuj Gulati, and Karunesh Ganguly. 2019. Activation of G-Protein Coupled Receptors.” Neuroscience “Competing Roles of Slow Oscillations and Delta Waves in Research 45 (4): 363–67. Memory Consolidation versus Forgetting.” Cell 179 (2): https://doi.org/10.1016/s0168-0102(03)00002-6. 514-526.e13. https://doi.org/10.1016/j.cell.2019.08.040. [26] Nishiyama, Makoto, Kyonsoo Hong, Katsuhiko Mikoshiba, [45] Farooq, U., and G. Dragoi. 2019. “Emergence of Preconfigured Mu-ming Poo, and Kunio Kato. 2000. “Calcium Stores Regulate and Plastic Time-Compressed Sequences in Early Postnatal the Polarity and Input Specificity of Synaptic Modification.” Nature Development.” Science 363 (6423): 168–73. 408 (6812): 584–88. https://doi.org/10.1038/35046067. https://doi.org/10.1126/science.aav0502. [27] Chen, Jiadong, Zhibing Tan, Li Zeng, Xiaoxing Zhang, You He, [46] Fernández-Ruiz, Antonio, Azahara Oliva, Eliezyer Fermino de Wei Gao, Xiumei Wu, et al. 2012. “Heterosynaptic Long-Term Oliveira, Florbela Rocha-Almeida, David Tingley, and György Depression Mediated by ATP Released from Astrocytes.” Glia 61 Buzsáki. 2019. “Long-Duration Hippocampal Sharp Wave Ripples (2): 178–91. https://doi.org/10.1002/glia.22425. Improve Memory.” Science 364 (6445): 1082–86. [28] SUL, JAI-YOON, GEORGE OROSZ, RICHARD S. GIVENS, and https://doi.org/10.1126/science.aax0758. PHILIP G. HAYDON. 2004. “Astrocytic Connectivity in the [47] Hara, Koji, and R. Adron Harris. 2002. “The Anesthetic Hippocampus.” Neuron Glia Biology 1 (1): 3–11. Mechanism of Urethane: The Effects on Neurotransmitter-Gated https://doi.org/10.1017/s1740925x04000031. Ion Channels.” Anesthesia & Analgesia 94 (2): 313–18. [29] Roth, Richard H., Robert H. Cudmore, Han L. Tan, Ingie Hong, https://doi.org/10.1097/00000539-200202000-00015. Yong Zhang, and Richard L. Huganir. 2020. “Cortical Synaptic [48] Feng, X., J.J. Liu, X. Zhou, F.H. Song, X.Y. Yang, X.S. Chen, W.Q. AMPA Receptor Plasticity during Motor Learning.” Neuron 105 (5): Huang, L.H. Zhou, and J.H. Ye. 2012. “Single Sevoflurane 895-908.e5. https://doi.org/10.1016/j.neuron.2019.12.005. Exposure Decreases Neuronal Nitric Oxide Synthase Levels in the [30] Dittman, Jeremy S., Anatol C. Kreitzer, and Wade G. Regehr. Hippocampus of Developing Rats.” British Journal of Anaesthesia 2000. “Interplay between Facilitation, Depression, and Residual 109 (2): 225–33. https://doi.org/10.1093/bja/aes121. Calcium at Three Presynaptic Terminals.” The Journal of [49] Vahle-Hinz, C, and O Detsch. 2002. “What Can in Vivo Neuroscience 20 (4): 1374–85. Electrophysiology in Animal Models Tell Us about Mechanisms of https://doi.org/10.1523/jneurosci.20-04-01374.200 Anaesthesia?” British Journal of Anaesthesia 89 (1): 123–42. [31] Tomar, Anupratap, Denis Polygalov, Sumantra Chattarji, and https://doi.org/10.1093/bja/aef166. Thomas J. McHugh. 2021. “Stress Enhances Hippocampal [50] Hentschke, H., A. Raz, B.M. Krause, C.A. Murphy, and M.I. Banks. Neuronal Synchrony and Alters Ripple-Spike Interaction.” 2017. “Disruption of Cortical Network Activity by the General Neurobiology of Stress 14 (May): 100327. Anaesthetic Isoflurane.” British Journal of Anaesthesia 119 (4): https://doi.org/10.1016/j.ynstr.2021.100327. 685–96. https://doi.org/10.1093/bja/aex199. [32] Ohtsuki, Gen. 2019. “Modification of Synaptic-Input Clustering by [51] Bratuša, M. 2022. “LTP and LTD dependence on spontaneous Intrinsic Excitability Plasticity on Cerebellar Purkinje Cell activity in glutamate synapses of the HPC and CTX and the role of Dendrites.” The Journal of Neuroscience 40 (2): 267–82. anaesthetics.” exchange semester project MEi:CogSci. https://doi.org/10.1523/jneurosci.3211-18.2019. https://www.academia.edu/80928790/LTP_and_LTD_dependence [33] Emptage, Nigel J., Christopher A. Reid, and Alan Fine. 2001. _on_spontaneous_activity_in_glutamate_synapses_of_the_HPC_ “Calcium Stores in Hippocampal Synaptic Boutons Mediate and_CTX_and_the_role_of_anaesthetics Short-Term Plasticity, Store-Operated Ca2+ Entry, and 13 Trusted sources and disinformation: studying the limits of science* Rita Gsenger Weizenbaum Institute, Humboldt University Berlin Berlin, Germany rita.gsenger@hu-berlin.de ABSTRACT Disinformation often speaks to human emotions and touches upon controversial or ideologically charged topics [5] as “people During the Covid-19 pandemic, the spread of disinformation have a taste, a predisposition even, for it” [5, p. 57], and virality became more apparent. Much of that disinformation focused on helps content to be distributed widely [6]. As soon as such topics health-related topics and the current health crisis, often claiming are concerned, the content is often not that important to be to be scientific information. Trusting scientists became crucial to empirically correct and reasonable. However, a social identity is counter the pandemic effectively as a society; however, science-afforded by believing that content is valuable [7]. Making certain related disinformation and so-called pseudoscience provided beliefs their identity does not only lead to an ignorance of facts new challenges for societies. These beliefs often overlap with (ibid.), but it also enables people to think along ideologically other types of disinformation and conspirational thinking, polarised lines with affective disdain for outgroup beliefs [8]. making them very attractive to human cognition. Twenty semi- Therefore, believing, for instance, that the earth is flat goes structured interviews were done in 2020 to investigate beyond holding a belief but is used to form identities. Around individuals' trust in science, governments, and media. The these identities, movements are formed, and, in the case of flat interviews focused on information sources and the conclusions earthers, for instance, people are willing to lose their jobs, drawn from the situation to determine how individuals estimate friends, and family relations to be part of the group [7]. information sources' trustworthiness. A kind of disinformation that has become crucial, especially during the Covid-19 pandemic, has been dubbed pseudoscience. KEYWORDS Pseudoscience, disinformation, Covid-19, trust 1.2 What is pseudoscience? 1 DISINFORMATION AND PSEUDOSCIENCE Distinguishing science from pseudoscience is not a simple endeavour. Some paradigmatic cases might exist where 1.1 Dimensions of disinformation philosophers and scientists agree, but other examples remain undecided or on the fringes of science. Ultimately, the question Disinformation is most commonly defined as false information of defining science and delimitating it from non-science comes that is deliberately propagated and distributed [1, 2, 3, 4]. The down to a fundamental question of philosophy, mainly what concept of disinformation includes various dimensions and knowledge is and how we attain it [9]. Pseudoscience can be aspects, which often overlap and influence each other [1, 3]. understood as a discourse about a specific subject matter, and Kapantai et al. (2020) developed a taxonomical framework to what is considered pseudoscience, like science, changes [10]. include important types of disinformation, including the motive Defining pseudoscience often “involves subjects that are either (profit, ideological, psychological, and unclear), facticity, and on the margins or borderlands of science and are not yet proven, verifiability as dimensions. That resulted in eleven kinds of or have been disproved, or make claims that sound scientific but disinformation, including, for instance, conspiracy theories, in fact have no relationship to science” [11, p. 203]. Several pseudoscience, hoaxes, trolling, or clickbait. Disinformation can characteristics can be identified to designate the likelihood of also be partly true to make it more credible [5]. adherence to pseudoscientific or non-scientific claims. For example, the language used to describe the phenomenon or ∗This abstract is partly based on the author’s Master thesis: Digital Literacy and Pseudoscience in Crisis Response. The Case of COVID-19 in Austria (University of research results often indicates the credibility of the reported Vienna 2021). results or evidence. The excessive use of technical terms and scientifically sounding language, for instance, in press releases, Permission to make digital or hard copies of part or all of this work for personal or classroom use is granted without fee provided that copies are not made or distributed might lead to trust and acceptance of the presented results due to for profit or commercial advantage and that copies bear this notice and the full the impression of smart people doing important work. These are, citation on the first page. Copyrights for third-party components of this work must be honored. For all other uses, contact the owner/author(s). however, not doing a great job in communicating science, and Information Society 2022, 10–14 October 2022, Ljubljana, Slovenia © 2022 Copyright held by the owner/author(s). 14 Information Society 2022, 10–14 October 2022, Ljubljana, Slovenia Rita Gsenger more importantly, some elements of good scientific practice can 2.2 Results: “I don’t know what to believe be commonly understood [12]. Pseudoscientific theories often anymore”: doubt and trust in times of crises use language full of epithets and refer to emotions and religion or use ideological markers.The presentations often include theses Participants used various sources of information about the and evaluations presented as unequivocal [13]. The method used Covid-19 pandemic, including online sources, TV, radio and might not be scientifically sound. For instance, anecdotal conversations with friends and family. When asked about the evidence and not controlling for other variables, very small or sources participants considered trustworthy or not to provide unrepresentative sample sizes in establishing causal information about the Covid-19 pandemic, various categories relationships, lack of control groups, or blind testing might were mentioned, including the media, social media, social indicate unsound methods. Moreover, many pseudoscientific contacts, and the government. However, the trust did not seem to studies tend to select parts of their evidence, which allows for a be easily acquired or granted among participants. Social aspects very charitable interpretation of studies to support a predefined were considered influential in attributing trustworthiness (P3, conclusion [12]. As there is not one comprehensive definition of P6). Therefore, a reason to trust a source would be that people pseudoscience, issues fall more or less under its spectrum. I will from an individual’s social circle would also trust it (P3, P6, P19). consider the above-outlined characteristics during my empirical Furthermore, authenticity and “thinking outside the box” (P13) investigation. Pseudoscience and science are historical were considered trustworthy traits of people. Some said they phenomena that inform the decisions societies make about what would trust family members, doctors or journalists they knew is considered the truth. Attempting to define pseudoscience personally (P2, P11). involves making claims about the nature of science. Overall, no methodology has been developed that allows for a general and comprehensive distinction [10], with Popper’s principle of Participants trust media if they provide sources with more falsifiability [14] not solving the problem satisfactorily [15]. In a information about the topic in question (P3), including links to culture that highly values science, other domains such as religion, other trusted websites (P4). Furthermore, if the information could politics, or literature are often closely associated with science and be cross-referenced with scientific sources (P14), and if scientists, seem to borrow scientific language, theories, or methods [11]. experts, or studies are included (P20), the trust in media sources Later, theories might be reevaluated and reclassified as science is increased. Furthermore, including various opinions was or pseudoscience [10]. Pseudoscientific beliefs are not a considered a sign of trustworthiness (P6, P1). These opinions marginal phenomenon and influence public policies [9, 16]. For permit looking at a subject from multiple viewpoints (P3) and example, during the Covid- 19 pandemic, pseudoscientific discussions by different people (P3, P14). The content would not explanations for the causes and cures of the virus surged [17]. be considered trustworthy if a personal opinion were presented Therefore, these beliefs, especially in crises, when public as objective truth (P3). Furthermore, the presentation of policies might be more crucial to follow, and such beliefs could information in the media and on the internet influences the be detrimental to society. However, belief acquisition is not attribution of trustworthiness. Accordingly, the way people post always easy, as human beings are prone to biases and faulty something, specifically the language (P14), if they write whole conclusions. sentences and if they explain the context of an article (P4) or if something is not formulated blatantly (P18) and frequently based on emotion (P5) it is considered more trustworthy. However, 2 STUDYING THE LIMITS OF SCIENCE trust was not attributed without reservation for many participants as they perceived the media as having their agenda (P8, P13) and 2.1 Method and participants being prejudiced (P14), but still more trustworthy than social media (P8). On the other hand, some did not consider the media Twenty semi-structured [18] and problem-centred qualitative “a source to find out what is really happening” (P13), and one interviews [19] were conducted in November and December participant mentioned that they “don’t believe anything anymore” 2020 with Austrian volunteers (N=20, 16 female, age 19-65, SD because “[...] it is not explained what the numbers mean at all or = 13.8). Interviews were led in German, and the author translated put into a context from which area the numbers come from and quotes. Interview participants were volunteers. Therefore, the how they were created at all” (P16). Social media was not researcher did not have much influence on their gender. However, considered trustworthy because a lot of information originates gender was determined not to be a crucial influence on the study. from private individuals (P19). Moreover, assessing the The discussions included several topics. However, only one part trustworthiness of information on social media is challenging focusing on trust and attitude towards the government, scientists (P1), even though most participants considered some people they and media is contained in this paper. Some limitations must be were friends with on Facebook trustworthy (P3, P4, P15). Some outlined when doing qualitative interviews online. Conducting participants based their trust on intuition and how they felt interviews online limits the information transmission compared regarding the media and online information. One participant to real life interviewing face-to-face. The qualitative study was described it as follows: “When I open that, how does it ‘feel’ if I done at a specific moment of the pandemic and thus only reflects move towards a platform, then I read how the information is participants’ attitudes during that time. Furthermore, participants structured, and I read the first paragraphs, and when something might be hesitant to share pseudoscientific beliefs or denial of is in there that seems a bit strange to me, then I would get out of science with a researcher. Therefore, no outright questions about there and look it up somewhere else. So, it depends on how it is such ideas were asked. in a textual sense and how the information lies in front of me” 15 Trusted sources and disinformation Information Society 2022, 10-14 October 2022, Ljubljana, Slovenia (P15). More specific descriptions of that feeling included if Moreover, participants observed how people changed and something seemed “out of touch with reality” (P5), what sounds suddenly believed entirely different things (P16, P2, P8). reasonable (P3), to use one’s common sense or if it appears According to participants, everybody should state their opinion strange (P18). Participants furthermore attributed trustworthiness but has the responsibility to do it respectfully (P15). An to sources or information that would confirm a worldview. individual’s history is crucial to consider to make respectful Accordingly, a participant described that other people with interaction easier (P2). Participants elaborated more in detail on differing worldviews would find different information how they formed their own opinions about the measures, the trustworthy and objective (P6). Additionally, reputation was a communication, and the pandemic in general. Some attributed source of trustworthiness, especially in the media (P10, P19). the decision to believe the information from a source to intuition Some participants mentioned the government and ministries as (P12) or if it seems strange (P20), as highlighted previously in trustworthy sources of information. One participant said, “in the section 6.4.1. Furthermore, they highlighted the influence of last months, I have experienced things where I was not sure in social factors, such as the influence of people they would talk to the moment can I trust anyone, and this is now a purely emotional (P5), for instance, in their workplace or people who had the thing because you cannot know anything anyway” (P8). illness (P14), even if they disagreed with them (P5). They would Furthermore, a participant claimed that “somebody is telling me, like to discuss these issues among their circle of friends as some I cannot go to university anymore, that I cannot see people would be more active and critical and might introduce other anymore, who is that somebody who would permit that, who perspectives (P14). If some- thing seems strange, however, they decides about me, that I cannot do that anymore” (P5). Some would try to find other opinions (P20), and online they would participants showed understanding of the difficult decisions the follow links from friends (P9) or try and consume contrary government needs to take right now, claiming as they would not opinions (P5). Overall, participants would form their opinions by want to be in their position or get involved, they would need to combining various other opinions (P8), questioning their comply with measures (P11). worldview, and staying open for new information (P14), and reflect on it (P5). Participants highlighted difficulties with Furthermore, some participants claimed that everybody would opinion formation about the Covid-19 pandemic, as one need to find their way of dealing with the situation and meet as participant summarized: “I believe a big problem is that there are many people as they would think appropriate (P14), emphasizing so many people, where it is claimed, ok, I am a doctor in that area, the responsibility of individuals (P5, P14). Many participants and I say this and that. And the doctor then says that and you mentioned the adverse consequences of the measures. Some don’t know, is that person really a doctor, do they really know agreed that these consequences, including the dangers of a about that. I mean, probably they are doctors but did they actually lockdown (P1), were not discussed enough (P1, P2, P16, P17). engage with that issue, or are they just saying anything? There Some were worried about restricting civil rights during the are so many doctors that have different areas of expertise and, of lockdown and possible dangers to democracy (P9, P1), claiming course, various experiences and a different level of knowledge, that the government could not implement a curfew as it violated so you don’t know where the information is coming from” (P16). human rights (P1). Participants wished that people were given more credit (P20, P13), which included telling them to take care 3 DISCUSSION AND CONCLUSION of their immune system and take vitamins (P13). Another participant would have wanted different perspectives on the Participants highlighted some specific topics as instilling the transformation happening in 2020, as communication is most significant doubts about trustworthy sources considering changing and more telepathy will be possible due to that change the Covid-19 pandemic. Science was considered a trusted source, (P5). According to participant 13, not discussing alternative ways but various indications showed that participants had significant of handling the pandemic can be attributed to international doubts regarding the scientific consensus about Covid-19, for pressure (P13). instance, that they would not know what it really was (P1) and that it was the flu, which is unpleasant but not particularly The plurality of opinions is generally valued highly among dangerous (P13). Tests to determine infections were doubted in participants as it is essential that everybody can share their their validity and efficacy (P16) and are considered inaccurate standpoint and how they arrive at their conclusions because (P13). Even though the interviews were led before Covid-19 everybody has a good reason to think as they do (P2, P1). vaccinations were widely available, mandatory vaccinations However, according to some participants, not all opinions and were already a big concern for some participants, which are standpoints were listened to somewhat during the pandemic. For thought to change society (P11) and should be well prepared to instance, the questions “masks yes or no these questions are not take people’s fears about the vaccinations (P20) as chaos might allowed to be asked because we are being beaten down by all ensue if vaccinations become mandatory (P12). Furthermore, the these numbers” (P13), and they should listen to people who have topic of not being told everything was present regarding the issue other methods (P14). Some observed that a division between of vaccinations. Participants worried about what would happen opinions and people was taking place in the general society. In to the Austrian culture and the country if vaccinations were that regard, only two contrary camps seemed to exist, and only mandatory (P11). An electronic compulsory vaccination to “be for Corona or against, a middle course or a differentiated certificate was mentioned as a source of worry for a functioning account was not possible” (P17). That means, participants were peaceful democracy (P9). Various conspirational elements worried that a nuanced debate about issues regarding the seemed to be present during interviews. For example, some pandemic was more difficult. participants were worried about democracy and the rule of law in 16 Information Society 2022, 10–14 October 2022, Ljubljana, Slovenia Rita Gsenger Austria. One mentioned that a friend who is a doctor told them [3] E. Kapantai, A. Christopoulou, C. Berberidis and V. Peristeras, "A that now with the pandemic, the government can achieve things Systematic Literature Review on Disinformation: Toward a Unified they could never have done without the pandemic (P13), which Taxonomical Framework," New Media & Society, vol. 23, no. 5, pp. 1301-1326, 2021. happens behind the scenes and might endanger our democracy [4] K. Shu, S. Wang, D. Lee and H. Liu, "Mining Disinformation and Fake (P9). In that regard, with the climate of fear, the government “is News: Concepts, Methods, and Recent Advancements," in Disinformation, trying out how far it can tighten the thumbscrews” (P9), and the Misinformation, and Fake News in Social Media: Emerging Research government lies, meaning something changed in a significant Challenges and Opportunities, Cham, Springer International Publishing, 2020, pp. 1-20. and sustainable way without people knowing (P9). In that regard, a participant stated the government was “catholic, dishonest and [5] N. O'Shaughnessy, "From Disinformation to Fake News: Forwards into the Past," in The Sage Handbook of Propaganda, Thousand Oaks (CA), Sage, tendentious” (P11) and that it “actually does not have a plan [...] 2019, pp. 55-71. only false numbers, false facts, false something” (P17). Some [6] S. Vosoughi, D. Roy and S. Aral, "The Spread of True and False News doubted the democratic nature of the situation, as not everything Online," Science, vol. 359, no. 6380, pp. 1146-1151, 2018. is communicated (P18) in our “so-called democracy” (P12). One [7] L. McIntyre, How to Talk to a Science Denier. Conversations with Flat participant summarised the situation as follows: Earthers, Climate Deniers, and Others Who Defy Reason, Cambridge (MA): MIT Press, 2021. [8] L. Mason, "Ideologues without Issues: Polarizing Consequences of “Honestly, it is a bit authoritarian because the information comes Ideological Identities," Public Opinion Quarterly, vol. 82, no. S1, pp. 866-from above. Kurz [the Austrian chancellor] speaks from the 887, 2018. microphone, and everyone listens, sits in front of the TV or [9] M. Pigliucci and M. Boudry, "Introduction: Why the Demarcation Problem channel, listens to him, and then it is done. I don’t think that’s Matters," in Philosophy of Pseudoscience: Reconsidering the Demarcation Problem, London, The University of Chicao Press, 2013, pp. 1-9. democracy. How that has now developed individually that certain events were then possible, these self-initiatives that have [10] D. K. Hecht, "Pseudoscience and the Pursuit of Truth," in Pseudoscience. The Conspiracy Against Science, Cambridge (MA), MIT Press, 2018, pp. 3-then taken place in conformity with measures I find again thanks 20. to people with whom I live together that we are democratic. [11] M. Shermer, "Science and Pseudoscience. The Difference in Practice and the These two perspectives in my social environment where you get Difference It Makes," in Philosophy of Pseudoscience: Reconsidering the together and ask if it’s okay if you can hug someone or sing with Demarcation Problem. , London, The University of Chicago Press, 2013, pp. 203-223. each other even though choirs are not allowed to practice so that [12] C. W. Lack and J. Rousseau, Critical Thinking, Science, and Pseudoscience, in agreement with the others, of course, works because we are New York: Springer, 2016. not in the snitch system and the Biedermeier maybe it seems so [13] M. Szynkiewicz, "May you live in interesting times. Science vs. but not quite.” (P5). pseudoscience in the era of the internet," Ethics in Progress, vol. 11, no. 1, pp. 85-98, 2020. All participants seemed to struggle with doubts regarding [14] K. Popper, The logic of scientific discovery, London and New York: Routledge, 2002. handling the pandemic in 2020. These doubts focused either on [15] M. Pigliucci, "The Demarcation Problem. A (Belated) Response to the government or on science. The media seemed to be the most Laudan.," in Philosophy of Pseudoscience: Reconsidering the Demarcation trusted. However, some would argue that they would only report Problem, London, The University of Chicago Press, 2013, pp. 9-29. uncritically. If doubts focused on the government, they seemed [16] J. H. Taylor, R. A. Eve and F. B. Harrold, "Why creationists don't go to to lean more towards a conspirational mindset. Doubts regarding psychic fairs: differential sources of pseudoscientific beliefs," Skeptical scientific consensus about Covid-19 are mostly deemed to adhere Inquirer, vol. 19, no. 6, pp. 23-28, 1995. to pseudoscientific beliefs such as Covid-19 is the same as the [17] T. T. Desta and T. Mulugeta, "Living with COVID-19 triggered pseudoscience and conspiracies," International journal of Public Health, flu or tests/masks do not work. However, a mix of pseudoscience vol. 65, no. 6, pp. 713-714, 2020. (vaccines do not work) and conspirational tendencies (there is [18] G. D. McCracken, The long interview, Newbury Park (CA): Sage, 1988. some more extensive agenda) could be observed when discussing [19] U. Flick, An introduction to qualitative research, London: Sage, 2009. vaccinations. In conclusion, the frequency of social media use [20] R. Armitage and C. Vaccari, "Misinformation and Disinfomration," in The and the content consumed should not be overestimated, as an Routledge Companion to Media Disinformation and Populism, London and individuals’ immediate social environment (i.e. friends and New York, Routledge, 2021, pp. 38-49. family) seems to have a more significant influence on their [21] P. Carrillo-Santisteve and P. Lopalco, "Measles still spread in Europe: Who beliefs. is responsible for the failure to vaccinate?," Clinical Microbiology and Infection, vol. 18, no. 5, pp. 50-56, 2012. [22] J. M. Berman, Anti vaxxers: How to Challenge a Misinformed Movement, Cambridge (MA): Teh MIT Press, 2020. REFERENCES [23] D. Kahneman, "A perspective on judgement and choice: Mapping bounded rationality," American Psychologist, vol. 58, no. 9, pp. 697-720, 2003. [1] J. Möller, M. Hameleers and F. Ferreau, "Verschiedene Formen von [24] H. A. Simon, "Rational Decision Making in Business Organizations," The Desinformation und ihre Verbreitung aus American Economic Review, vol. 69, no. 4, pp. 493-513, 1979. kommunikationswissenschaftlicher und rechtswissenschaftlicher [25] A. Tversky and D. Kahneman, "Judgment under uncertainty: Heuristics and Perpektive. Ein Gutachten im Auftrag der Gremienvorsitzendenkonferenz biases," in Judgment under uncertainty: Heuritsics and biases, Cambridge der Landesmedienanstalten (GVK).," die medeinanstalten - ALM GbR, (MA), Cambridge Unviersity Press, 1982, pp. 3-23. Berlin, 2020. [26] A. Tversky, "Elimination by Aspects: A Theory of Choice," Psychological [2] E. Humprecht, F. Esser and P. Van Aelst, "Resilience to Online Review,, vol. 79, no. 4, pp. 281-299, 1972. Disinformation: A Framework for Cross-National Comparative Research," The INternational Journal of Press/Politics, vol. 25, no. 3, pp. 493-516, 2020. 17 Opacity and understanding in artificial neural networks: a philosophical perspective Martin Justin martin1123581321@gmail.com Faculty of Arts, University of Ljubljana Ljubljana, Slovenia ABSTRACT the 1980s talked about an explanatory task, the problem scientists’ face today could be called an explanatory barrier. In the paper, I review some of the emerging philosophical litera-In this paper, I will review some of the emerging philosophical ture on the problem of using artificial neural networks (ANNs) literature on the problem of using ANNs in science. First, I will and deep learning in science. Specifically, I focus on the problem briefly introduce the problem of opaqueness or the so-called back-of opacity in such systems and argue that although using deep box problem. Then, I will present a paper by Erasmus et al. [6] neural networks in cognitive science can produce better results, which provides a detailed analysis of the notions of explanation it can also acts as a barrier to gaining new understanding of and understanding that are central to thinking about the problem. cognitive processes. After that, I will present Florian J. Boge’s [3] argument that we KEYWORDS can talk about two distinct dimensions of opacity in ANNs. In the last section, Mazviita Chimirmuuta’s [5] argument about explanation, understanding, scientific discovery, artificial neural the implications of the trade-off between predictive accuracy networks, black-box problem and opacity for research in computational neuroscience will be presented. 1 INTRODUCTION Early on in their inception, connectionist approaches in cog- 2 BLACK BOX PROBLEM nitive science faced challenges from proponents of competing Let us first turn to the problem of opaqueness. Authors of one approaches. One of the leading theorists of the classical symbolic of the review papers [10] from the field of explainable AI (XAI) approach, J. Fodor and Z. Pylyshyn [7], for example, argued that note that the “predictive accuracy [of machine learning systems] connectionism could not account for four essential properties has often been achieved through increased model complexity.” of cognition – i.e., productivity, systematicity, compositionality, This increased complexity, “combined with the fact that vast and coherence – and thus was not a sufficient theory of the mind. amounts of data are used to train and develops such complex We now have good reasons to believe that their argument does systems” has inherently reduced researches’ ability to “explain not hold [11]. Indeed, in their demonstration of the supposed the inner workings and mechanisms” of these systems. As a inadequacy of connectionist models, Fodor and Pylyshyn only result, “the rationale behind decisions [of these systems] becomes considered very simple models with local representations. But it quite hard to understand and, therefore, their predictions hard turns out that more complex models with distributed represen-to interpret.” Therefore, they say that “there is clear trade-off tations can satisfactorily solve the explanatory task. Contrary between the performance of a machine learning model and its to what Fodor and Pylyshyn claimed, we can therefore show ability to produce explainable and interpretable predictions.” The that even connectionist cognitive models are powerful enough authors of anther review paper [1] reached a similar conclusion: to exhibit the required properties. “Indeed, there are algorithms that are more interpretable than Fast forward forty or so years in the future, scientists using others are, and there is often a tradeoff between accuracy and artificial neural networks (ANNs) and deep learning to study interpretability: the most accurate AI/ML models usually are cognitive functions now face a different problem. One of the not very explainable (for example, deep neural nets, boosted key advantages of present day ANNs that use deep learning is trees, random forests, and support vector machines), and the their increased complexity and depth [4]. But because of their most interpretable models usually are less accurate (for example, increased complexity, such systems can become opaque in a way linear or logistic regression).” that even the researchers developing them do not understand Authors of [10] thus distinguish between "black-box" models, some key aspects of how they work [10]. Present day ANNs can which have state-of-the-art performance but are opaque, and thus be used to model cognitive functions much more successfully "white-box" or "glass-box" models, which are more easily in-than before, but because of their opaqueness, it is unclear what terpretable, but not as powerful. In her paper, Chirmuuta [5] 1 new insights such successes are generating [5]. If researchers in also specifies which aspects of deep neural networks suffer from opaqueness. She argues that scientists have a good understanding 1 In contrast to this, Sullivan [14] argues that the problem of contemporary ANNs of “internal architecture and workings” of the systems, i.e., they is not their opacity, but “a lack of scientific and empirical evidence supporting the link that connects a model to the target phenomenon.” But see Räz and Breisbart know the activation values of the units, the learning rule, the depth of the network and the connectivity between the layers. Permission to make digital or hard copies of part or all of this work for personal But they do not know exactly how an already trained network or classroom use is granted without fee provided that copies are not made or arrives at a prediction or classification. distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. Copyrights for third-party components of this work must be honored. For all other uses, contact the owner /author(s). Information Society 2022, 10–14 October 2022, Ljubljana, Slovenia © 2022 Copyright held by the owner/author(s). [13] for an argument that her point rests on a weak and thus undesirable notion of understanding. 18 Information Society 2022, 10–14 October 2022, Ljubljana, Slovenia Martin Justin 3 EXPLANATION AND UNDERSTANDING world”, and thus involves giving information about the causal process and the causal interaction that leads to the emergence The black box problem or the problem of opaqueness has re-of the explanandum. (IV) New Mechanist model which takes as sulted in increased attention to research in explainable AI. But explanans the entities and their activities that are responsible for one salient feature of the literature on explainable AI is the impre- 3 the emergence of the explanandum. cise or even interchangeable use of the concepts of explainability, Erasmus et al. [6] then argue (b) that the increased complexity intelligibility and interpretability. This is also recognized by the of the phenomenon we are trying to explain (or of the concepts researchers themselves. For example, the authors [10] observe and data we use to explain it) does not affect our ability to offer that there is “no concrete mathematical definition of the concepts an explanation for the phenomenon. And (c) that deep neural net-of explainability and interpretability.” Nevertheless, they make a works can be explained in all four of the ways described above. conceptual distinction between these two terms. Interpretability, The argument for (b) is quite simple: Deductive Nomological on the one hand, is understood in connection to the ability of explanation, for example, requires only that the explanans con-researchers to intuitively understand the relationship between tains a law, and that the process of explanation takes the form inputs and outputs of a system. Explainability, on the other hand, of deductive reasoning. It does not matter how complex the two is understood in relation to the ability to understand the inner elements are. Thus, an explanation that contains a more complex workings of a system. In contrast, authors of a different similar explanans and requires more complex reasoning may be less study [9] define explainability as possibility to provide a satis-desirable, but it is no less an explanation. factory answer to the "why" question regarding the functioning The argument for (c) is a bit more technical. To demonstrate of a system. They also make a difference between two levels of this point, the authors provide an example of an explanation of explanation, connected to two different questions scientists can how an ANN, trained to identify dense breast tissue on X-ray ask about a system. Namely, “why does this particular input lead images, classify these images [6]. Let us see how a Deductive to that particular output?” i.e., a question about the relationship Nomological explanations of such ANN could work. As the em-between inputs and outputs, and “what information does the pirical content of the explanans, we could use all the information network contain?” i.e., a question about the internal workings of about the activation values of the individual units in the network a system. and about the weights between them, as well as the numerical In their paper, Erasmus et al. [6] point to this shortcoming values of the input data. We could also form a law-like proposi-of the literature on explainable AI and argue that this imprecise tion of the form “outputs with such and such numerical value use of the terms leads to a misunderstanding of the trade-off are classified as such and such.” In this way, the explanandum, between performance and explainability of AI systems. Their i.e., the classification of the photograph F into the class r, would argument proceeds in three steps. First, they offer a more pre-be explained using an explanans consisting of a law-like prepo-cise analysis of the notions of explanation and understanding. sition and empirical content. In other words, we would have a Second, they show that the increased complexity of systems af- 2 Deductive Nomological explanation. Although the arguments for fects their undersandability rather than their explainability. And (b) and (c) were presented only for the case of Deductive Nomo-third, they offer a typology of possible explanatory methods that logical explanation, authors argue they apply mutatis mutandis could also increase the intelligibility of systems. Here, I will be to other models of explanation as well. interested mainly in the first and the second step. Therefore, in Let us now turn to (d), the definition of understanding. As the remainder of this section I will first present (a) their defini-Erasmus et al. [6] point out, authors who study understanding tion of explanation, (b) their arguments that the possibility of do not, of course, entirely agree on its exact definition, but they explanation is independent of the complexity of the phenomenon commonly observe that, while explanation is necessary for un-itself, and (c) the argument that ANNs can be explained. Then derstanding, it is not sufficient for it. So to gain understanding I will present (d) their definition of understanding and (e) their of a phenomenon, some other conditions besides having an ex-argument that complexity affects the ability to understand. planation must be met. There are several candidates for these Let us start with (a). In defining the notion of explanation, additional conditions in the literature, but, as Erasmus et al. argue, Erasmus et al. [6] draw on a longer tradition in philosophy of they all have in common that they are “psychological traits of the science which holds that explanation consists of three elements: user of the explanation.” One such condition is the criterion of (1) the explanandum, i.e., what we want to explain, (2) the explanans intelligibility. It states that a theory T is intelligible to a scientist , i.e., with what we are explaining, and (3) the process of explanation in a context C if the scientist is able to recognize the qualitatively . Different models of explanations differ in one or distinct consequences of T without doing the exact calculations more of these elements. Four such models feature prominently 4 [5, 6]. Given this, it is obvious that increased complexity of an in the literature. (I) Deductive Nomological model, in which the explanation or a phenomenon makes it less intelligible and thus explanans includes empirical content plus a law-like preposition, less understandable. Thus, it can be concluded that (e) complexity and the process of explanation takes the form of deductive rea-affects the ability to understand. soning. (II) Inductive Statistical model, in which the explanans includes a statistical law about behavior of the variables, and 4 TWO DIMENSIONS OF OPACITY the process of explanation takes the form of inductive or probabilistic reasoning. (III) Causal Mechanical model which aims to Erasmus et al. [6] argue that while the workings of deep neural show “how the explanandum fits into the causal structure of the networks are explainable, they are often not understandable for 2 3 See Beisbart and Räz [2] for a critique of this point. They say that “the distinction Woodword and Ross [17] present a slightly different typology. In particular, they that Erasmus et al. draw between interpretability and explainability in this way add Salmon’s statistical relevance model and pragmatic models of explanation. 4 seems rather stipulative.” In contrast, they argue that we should use these terms as Chirimuuta [5] also lists four properties of a theory (or an explanation) that affect synonyms. Nevertheless, I think that Erasmus et al. [6] point to an important and its intelligibility. Those are: (1) the possibility of visualization, (2) the simplicity of well established concetpual distinction between these two terms which should not included theoretical assumptions, (3) the linearity of mathematical operations, and be so easily dismissed. (4) functional transparency. 19 Opacity and understanding in ANNs Information Society 2022, 10–14 October 2022, Ljubljana, Slovenia human users. In other words, they conclude that we should talk using a w-opaque deep neural network; (c) finally, I will present about a trade-off between the performance of AI systems and Chirimuuta’s version of the trade-off between performance and their understandability or intelligibility, not their explainability. understanding that arises when using ANNs in science. Nevertheless, they seem to overlook another important aspect Let us start with (a). Chirimuuta [5] defines computational of the trade-off. As it is apparent from the definitions of explain-neuroscience as “a tradition of research that builds mathematical ability and understandability in Gilpin et al. [9] and Linardatos models of neurons’ response profiles, aiming both at predictive et al. [10], there seem to be different ways in which ANNs can accuracy and at theoretical understanding of the computations be opaque to humans. performed by classes of neurons.” It is based on the assumption This point is explicated and extended upon by Boge [3]. In that information about the external world is ‘encoded’ in the elec-his paper, he presents the following three theses: (1) deep neural trical and chemical signals of the neurons. It attempts to solve the networks are instrumental, and their instrumentality is distinct so-called ‘decoding problem’, i.e., it tries to find a mathematical from that of other mathematical models; (2) deep neural networks function that could successfully link neuron spikes to outside are opaque in two different ways; and (3) the combination of (1) information. Specifically, according to Chirimuuta, scientists try and (2) means that in the future, we may not be able to understand to devise a theory of how neurons encode information about the potential new discoveries made by deep neural networks. In the outside world and then write a program, called an encoder, that rest of this section, I will be primarily interested in (2). performs the translation operation between the stimuli and the Boge [3] begins his exposition of the two aspects of opacity neural activity. by defining opacity. He defines it as follows: “a process P is Thus, as Chirimuuta [5] points out, computational neuro-epistemologically opaque to a subject X at time t if and only if science pursues two separate epistemic goals. On the one hand, it X does not know all the epistemically relevant elements of the aims at accurately predicting the relations between neural activ-process P at time t.” He then distinguishes between two aspects of ity and external stimuli (e.g., to predict how neurons will fire if the opacity of deep neural networks. First, he describes h-opacity. we show a picture of a square to a primate). On the other hand, it It concerns the operation of a system: a system is h-opaque if tries to understand how this translation takes place. Chirimuuta it is the process of its operation that is not not intelligible to its thus argues that in the past, when even very simple linear models human users. This is the opacity that results from the complexity have proved surprisingly accurate in certain contexts, there has of deep neural networks and hiders the understanding of the been a convergence between these two goals. However, with the connection between input and output data. But as Boge notes, development of deep neural networks, which are much more this type of opacity is not qualitatively different from, say, the accurate but w-opaque, these two goals started to diverge. opacity of other complex computational simulations, e.g., climate Chirimuuta [5] presents two examples of such divergence, simulations. He therefore identifies another aspect of opacity one from modeling the functioning of the motor cortex and an-that is specific to deep neural networks. This is w-opacity, which other from modeling the visual perception system. I will limit my concerns the representational content of the system (what was presentation to the former, i.e., to her comparison between two learned). According to Boge, in deep neural networks, not just the studies that tried to model motor cortex activity, Georgopoulos process that takes a neural network from an input to an output, et al. [8] and Sussillo et al. [15]. In both of these two experiments, but also the properties of the input data that guide this process researchers measured the activity of individual neurons in non-are opaque. human primates while the primates were performing given tasks. This difference is important as it points to a specific problem Georgopoulos et al. [8] present an experiment in which a monkey that the use of deep neural networks introduces to scientific was surrounded by eight buttons, with another button straight research. H-opacity only hinders the understanding of the com- ahead. In the experiment, first the button in front of the monkey putational model itself, as it prevents researchers from seeing lit up. After the monkey held it for one second, one of the other how it gets from input to output data. Such opacity can thus be eight buttons lit up, and the monkey had to press it with the problematic form an ethical point of view, as it makes it harder same hand. Meanwhile, the scientists measured the activity of a to justify the decisions made on the basis of a recommendation population of neurons in her motor cortex, and tried to establish by an AI system. In contrast to this, w-opacity reduces the po-a correlation between this activity and the direction of her arm tential of deep neural networks to bring new understanding to movement. They did this by simply converting the activity of a the processes studied by the scientists. Even in the case where neuron into a vector in three-dimensional space according to a promising results would suggest that an ANN represents a given formula they devised, and then summing the vectors of the indi-problem space in a better way than existing theories, w-opacity vidual neuronal cells to obtain one vector that represented the would leave this representation incomprehensible to scientists. whole neuron population. They found out that the direction of Thus, w-opacity has important implications for the use of neural this vector quite closely matched the direction of arm movement. networks in scientific research. Because of the fairly simple math they used, their model was completely transparent. In addition, the researchers themselves determined which information about the neural activity is im-5 PREDICTION VERSUS UNDERSTANDING portant and should be used to calculate the movement vector. The implications of w-opacity for research in computational neu-The accuracy achieved by the model can thus be seen as a partial roscience are convincingly presented by Chirimuuta [5]. In this confirmation that these features of neural activity are indeed section, I will summarize her findings. I will do this in the fol-important for directing arm movement. lowing steps: (a) first, I will briefly outline the research program The experiment reported by Sussillo et al. [15] is a bit different. of computational neuroscience; (b) then, I will present exam- They also had non-human primates, this time two, implanted ples of two studies from the field, one in which scientists ap-with electrodes that measured the activity of individual neurons proached their problem using a transparent mathematical model, in their motor cortex. But the monkeys did not press buttons; and another in which they approached a very similar problem rather, they had to move a cursor on a screen from a central 20 Information Society 2022, 10–14 October 2022, Ljubljana, Slovenia Martin Justin position to a marked position in one of the corners of the screen. shown in more detail what philosophers mean when they talk Each monkey performed three series of experiments. First, they about the trade-off between performance and intelligibility (or moved the cursor by moving their hand. Then, they moved the understandability) of AI systems in science. cursor using a brain-machine interface (BMI) that used a encoder, based on a mathematical model, similar to the one described in ACKNOWLEDGMENTS the previous example. In the last series, they used a BMI that en-I would like to thank Olga Markič for her useful suggestions and coded information using a trained neural network. Each monkey encouragement. I would also like to thank Nejc for his help with performed each of the three experiments hundreds of times. The the more technical aspects of the literature. researchers found that using this ANN based encoder significanty improved monkey’s performance vis-à-vis the older model. This REFERENCES suggests that the BMI with an ANN was more successful in trans- [1] Amina Adadi and Mohammed Berrada. 2018. Peeking Inside the Black-Box: lating between neuronal activation and information about the A Survey on Explainable Artificial Intelligence (XAI). IEEE Access, 6, 52138– 52160. doi: 10.1109/ACCESS.2018.2870052. outside world. We can thus assume that the ANN either approxi- [2] Claus Beisbart and Tim Räz. 2022. Philosophy of science at sea: Clarifying mated the mathematical function linking neuronal activation and the interpretability of machine learning. Philosophy Compass, 17, 6, (June 2022). doi: 10.1111/phc3.12830. external stimuli more accurately or it ‘discovered’ new properties [3] Florian J. Boge. 2021. Two Dimensions of Opacity and the Deep Learning of the input data that play an important role in the translation. Predicament. Minds and Machines, (Sept. 2021). doi: 10.1007/s11023- 021- 09 But the ANN used was both h-opaque and w-opaque so despite 569- 4. [4] Cameron Buckner. 2019. Deep learning: A philosophical introduction. Phi-its improved performance, it did not provide scientists with a losophy Compass, 14, 10, (Oct. 2019). doi: 10.1111/phc3.12625. better understanding of how the motor cortex works. [5] M. Chirimuuta. 2021. Prediction versus understanding in computationally Turning to point (c), it should now be clear what Chirimuuta enhanced neuroscience. Synthese, 199, 1-2, (Dec. 2021), 767–790. doi: 10.100 7/s11229- 020- 02713- 0. [5] is getting at when she says that the use of ANNs creates [6] Adrian Erasmus, Tyler D. P. Brunet, and Eyal Fisher. 2021. What is In-a divergence between the goals of predictive accuracy and un- terpretability? Philosophy & Technology, 34, 4, (Dec. 2021), 833–862. doi: 10.1007/s13347- 020- 00435- 2. derstanding of neurological processes. Chirimuuta calls this di- [7] Jerry A. Fodor and Zenon W. Pylyshyn. 1988. Connectionism and cognitive vergence a trade-off between understanding and accuracy. The architecture: A critical analysis. Cognition, 28, 1-2, (Mar. 1988), 3–71. doi: trade-off arises because a problem can either be tackled with 10.1016/0010- 0277(88)90031- 5. [8] Apostolos P. Georgopoulos, Andrew B. Schwartz, and Ronald E. Kettner. models that are not the most accurate, but can be interpreted 1986. Neuronal Population Coding of Movement Direction. Science, 233, and can thus provide us with new understanding of the problem. 4771, 1416–1419. Or it can be tackled with ANNs, which, although they achieve [9] Leilani H. Gilpin, David Bau, Ben Z. Yuan, Ayesha Bajwa, Michael Specter, and Lalana Kagal. 2018. Explaining Explanations: An Overview of Inter-greater accuracy, are opaque and therefore do not bring new pretability of Machine Learning. In 2018 IEEE 5th International Conference understanding to scientists. on Data Science and Advanced Analytics (DSAA). IEEE, (Oct. 2018), 80–89. isbn: 978-1-5386-5090-5. doi: 10.1109/DSAA.2018.00018. In addition to presenting a dilemma from the point of view of [10] Pantelis Linardatos, Vasilis Papastefanopoulos, and Sotiris Kotsiantis. 2020. epistemic goals of science, the trade-off also has some practical Explainable AI: A Review of Machine Learning Interpretability Methods. implications. For example, increasing reliance on ANNs to an- Entropy, 23, 1, (Dec. 2020), 18. doi: 10.3390/e23010018. [11] Olga Markič. 2011. Kognitivna znanost: filozofska vprašanja. Aristej, Maribor. alyze data may be linked to issues related to trust in scientific [12] Zubin Master and David B. Resnik. 2013. Hype and Public Trust in Science. findings. In his influential analysis of epistemic trust, T. Wilholt Science and Engineering Ethics, 19, 2, (June 2013), 321–335. doi: 10.1007/s119 [16] argued that the reliance between the members of a scien- 48- 011- 9327- 6. [13] Tim Räz and Claus Beisbart. 2022. The Importance of Understanding Deep tific community is based on the “assumption that the results Learning. Erkenntnis, (Aug. 2022). doi: 10.1007/s10670- 022- 00605- y. [the scientists are] relying upon were arrived at by means of [14] Emily Sullivan. 2022. Understanding from Machine Learning Models. The British Journal for the Philosophy of Science, 73, 1, (Mar. 2022), 109–133. doi: professional methods suitably employed”. Given the opacity of 10.1093/bjps/axz035. ANNs using deep learning, this assumption might be difficult [15] David Sussillo, Paul Nuyujukian, Joline M Fan, Jonathan C Kao, Sergey to test. Furthermore, some researchers speculated that “hyping” D Stavisky, Stephen Ryu, and Krishna Shenoy. 2012. A recurrent neural network for closed-loop intracortical brain–machine interface decoders. scientific results (especially in the more directly applicative fields, Journal of Neural Engineering, 9, 2, (Apr. 2012), 026027. doi: 10.1088/1741-2 such as biotechnology) can ultimately result in a loss of public 560/9/2/026027. trust in science. Although this connection between hype and [16] Torsten Wilholt. 2013. Epistemic Trust in Science. The British Journal for the Philosophy of Science, 64, 2, (June 2013), 233–253. doi: 10.1093/bjps/axs007. public trust have not yet been empirically established [12], it [17] James Woodward and Lauren Ross. 2021. Scientific Explanation. (2021). is not hard to see how focusing on predictive accuracy, rather Retrieved Jan. 29, 2022 from https://plato.stanf ord.edu/entries/scientif ic- ex planation/. than understanding, could further increase the unwanted hype surrounding scientific research. 6 CONCLUSION In this paper, I reviewed some of the emerging literature on the epistemological aspects of the problem of opaqueness in deep neural networks. First, I used Erasmus et al. [6] to point out that we need to distinguish between explainability and understandability of AI systems. I also presented their argument that the increasing complexity of these systems has a particular impact on our ability to understand them, not on their inherent explanaibil-ity. Then, with the help of Boge [3], I distinguished between two dimensions of opacity of these systems. Finally, following Chirimuuta [5], I presented this problem using a concrete example of two studies in computational neuroscience. In this way, I have 21 Politizirana znanost in zaupanje v znanost kot politična uniforma Politicized science and trust in science as a political uniform Jar Žiga Marušič† Famnit, Univerza na Primorskem Oddelek za psihologijo Koper, Slovenija jar.marusic@famnit.upr.si POVZETEK of trust in science in practice – endorsement of the process of scientific research, or an expression of political conformity? Zaupanje v znanost je dandanes, sploh po dveh letih pandemije Covida-19, posebej družbeno relevanten problem. Izraz pa je KEYWORDS nekoliko dvoumen, saj lahko znanost razumemo na več načinov, med drugim kot raziskovalni proces in kot institucije, na katerih Politicization of science, Covid-19, trust in science, social se ta proces odvija. Zato je izraz zaupanje v znanost lebdeči reasoning, psychological inoculation označevalec, oznaka brez jasnega referentnega objekta.. Težava lebdečih označevalcev se pokaže, ko postanejo tarča politizacije. 1 Kaj pomeni zaupati v znanost? V tem primeru zaradi nejasnosti semantičnega pomena sociopolitične konotacije izraza postanejo njegov primarni Vse večjo relevantnost pojma “zaupanja v znanost” v sodobni pomen. V politizirani znanosti bi zato “zaupati v znanost” v družbi lahko jemljemo kot posledico poznanstvenjenja družbe in resnici pomenilo podpirati obstoječi politični režim, izražanje družbenih praks – vse večjega soodvisnost znanstveno-tega zaupanja (ali njegovega pomanjkanja) pa bi služilo kot tehnološkega razvoja in vodenja sodobnih družbenh praks [22]. politična uniforma, ki izraža pripadnost enemu ali drugemu Še posebej pomemben pa je postal v zadnjih dveh letih, odkar se političnemu polu. V prispevku analiziram znanstveni diskurz je svet soočil s pandemijo Covida-19. Narodne, mednarodne ali zadnjih dveh let z namenom ugotavljanja, kaj je bil v tem celo globalne zdravstvene krize, kamor spadajo tudi pandemije, obdobju družbeni pomen zaupanja v znanost – podpora procesa so pogosto zaznamovane z določeno mero vključevanja znanstvenega raziskovanja ali izraz politične konformnosti? medicinske znanosti v vodenje družbe in usmerjanje družabnega življenja in Covid kriza je bila še posebej izrazit primer tega. KLJUČNE BESEDE Tako smo bili priča vsesplošni uporabi slogana “zaupajmo v znanost” (včasih “zaupajmo znanosti”), v angleščini “trust the Politizacija znanosti, Covid-19, zaupanje v znanost, socialno science” z namenom upravičevanja in izpostavljanja znanstveno presojanje, psihološka inokulacija podprtega značaja uradno sprejetih ukrepov za spopadanje s ABSTRACT Covid epidemijo. Kaj natanko pomeni zaupati v znanost? Drugače povedano, Trust in science is especially relevant in today’s society, given kateri znanosti naj bi se zaupalo? Znanost lahko razumemo kot that we are living in the wake of the 2-year Covid-19 pandemic. metodo (znanstvena metoda), proces (znanstveno-raziskovalni The term itself is somewhat vague, as science has multiple proces), socialni sistem (skupnost znanstvenikov) ali institucijo definitions, mainly the process of scientific research as well as (skupek akademskih institucij, kjer se izvaja znanstveno the institutions that engage in said process. Thus, trust in science raziskovanje). Vidimo torej, da pojem zaupanja v znanost nima is a floating signifier, a label without a clear referent. Such labels enoznačnega pomena – lahko pomeni zaupati kateri koli can be problematic if targeted by politicization. The vagueness kombinaciji zgoraj naštetih vidikov znanosti – zato ga je tudi of the floating signifier’s semantic meaning allows the socio-težko enoznačno vrednotiti in proučevati. Zaupanje v znanost je political connotations to acquire primacy. In times of politicized Hackingova človeška vrsta, je posplošitev oziroma klasifikacija science, “trusting in science” would then actually mean to neke človeške lastnosti oziroma vedenjske tendence, ki v svoji endorse the established political regime. As for actions that prisotnosti ali odsotnosti definira posebno kategorijo človeka signal this trust (or lack thereof), they would act as a political [10]. Človeške vrste so podvržene učinku zanke, zaradi refleksije uniform – an expression of political allegiance to one’s chosen in samo-refleksije identifikacija neke socialne entitete z side. This article analyses the state of scientific discourse during določeno človeško vrsto vpliva na lastnosti te socialne entitete, the pandemic, with the goal of establishing the precise meaning kar posledično vpliva tudi na pomen človeške vrste – oznake, s Permission to make digital or hard copies of part or all of this work for personal or katero jo poimenujemo. Pomen besede “znanost” je relativen in classroom use is granted without fee provided that copies are not made or distributed dinamičen tudi v odsotnosti zankanja, zato to velja tudi za for profit or commercial advantage and that copies bear this notice and the full citation on the first page. Copyrights for third-party components of this work must “zaupanje v znanost” – ko se spreminja pomen znanosti, se be honored. For all other uses, contact the owner/author(s). spreminja tudi pojem »zaupanje v znanost«. Information Society 2022, 10–14 October 2022, Ljubljana, Slovenia Zato lahko trdimo, da je zaupanje v znanost lebdeči © 2022 Copyright held by the owner/author(s). označevalec (ang. floating signifier), oznaka brez točnega ali splošno-sprejetega pomena, torej brez točnega referentnega 22 objekta [13]. Ravno v tej značilnosti se skriva moč lebdečih za osebo, ki presoja. Če je politika razločevanje med prijatelji in označevalcev – nejasnost njegovega pomena dopušča sovražniki, potem je znanost politična kadarkoli primarni kriterij individualno konstrukcijo pomena. Tako je točen pomen za razločanje med znanstveno in neznanstveno trditvijo ni lebdečega označevalca relativen – za eno osebo ali skupino ljudi kvaliteta argumentacije in podprtost z dokazi, temveč status pomeni nekaj, za drugo nekaj drugega. njenega sporočevalca. Z drugimi besedami, dihotomija Zaupanje v znanost torej nima enoznačnega pomena, kljub prijatelj/sovražnik se v znanosti odraža, ko je “kdo je to rekel?” temu pa lahko to idejo ovrednotimo na podlagi različnih možnih pomembnejše vprašanje od “kako je bila izjava argumentirana?”. definicij. Najprej si zamislimo dva ekstrema, znanstveni V podrobnosti argumentacije se morda ne moremo popolnoma dogmatizem in radikalni skepticizem do znanosti. Dogmatik bo spustiti, lahko pa vsaj presodimo ali je argumentacija formalno-najverjetneje trdil, da smo dolžni zaupati vsem aspektom logično ustrezna. znanosti – v uporabnost znanstvene metode, zanesljivost Lebdeči označevalci so zaradi svoje nejasnosti in dvoumnosti znanstveno-raziskovalnega procesa pri odgovarjanju na idealna tarča za politizacijo. Politizirana oznaka poleg svojega raziskovalna vprašanja, verodostojnost znanstvenikov in semantičnega pomena dobi še sociopolitični pomen – prisotnost nevtralnost oziroma apolitičnost znanstvenih institucij. Radikalni referentnega objekta označuje prijatelja ali sovražnika (režima) skeptik, v kolikor njegova pozicija ne temelji na a-priornem oziroma pripadnika ingrupe ali outgrupe. Ravno zaradi zavračanju, pa se bo najbrž skliceval na uvide Foucaulta [8] in nejasnosti semantičnega pomena (oznaka pomeni različne stvari Lyotarda [14], ki sta izpostavljala neko mero relativnosti različnim skupinam) sociopolitični pomen nadvlada znanstvenega spoznanja. Posledično bo trdil, da znanstvene semantičnega in postane primarni. Tako potem lebdeči institucije niso apolitične, znanstveniki niso racionalni in zato označevalec postane univerzalna oznaka za sovražnika režima – niti verodostojni, znanstveno-raziskovalni proces in znanstvena točen semantični pomen besede sicer vsak razume po svoje, metoda pa nista univerzalno orodje za dostopanje do resnice, njena čustvena in moralna valenca pa sta enoznačni. Znanost je temveč orodje za perpetuacijo specifične jezikovne igre. v naši družbi pozitivna, torej bi primeru politizacije “zaupanje v Na srečo lahko uberemo vmesno pot, ki ustreza klasični znanost” v svoji lebdeči obliki označevalo pripadnike ingrupe koncepciji razsvetljenske znanosti in temelji na egalitarnem oziroma prijatelje režima, njegova odsotnost pa njegove odnosu do znanja in zavračanju dolžnosti laika, da zaupa sovražnike oziroma pripadnike outgrupe. intelektualni avtoriteti. To stališče dobro povzame izjava S tega vidika je bila Covid kriza zelo zanimiva. Moja analiza Richarda Feynmana, da je znanost verjetje v nevednost se bo sicer osredotočala predvsem na dogajanje v mednarodni in strokovnjakov [7]. Potemtakem “zaupanje v znanost” pomeni ameriški znanosti, vendar so bili enaki ali podobni vzorci prisotni priznavanje uporabnosti znanstvene metode in zanesljivosti tudi v Sloveniji. Ekipa znanstvenikov iz MIT-ja je leta 2021 raziskovalnega procesa, hkrati pa ohranitev zdravega dvoma v objavila pre-print študije Viral Visualizations: How Coronavirus verodostojnost znanstvenikov in institucij. Če strokovnjak ali Skeptics Use Orthodox Data Practices to Promote Unorthodox institucija trdi da p, ni potrebno da temu slepo verjamemo, Science Online, ki je poročala o navadah, značilnostih, stališčih temveč lahko zahtevamo argumentacijo in vpogled v in vrednotah spletnih skupnosti Covid-skeptikov oziroma anti-raziskovalni proces. maskerjev, ljudi, ki so tako ali drugače nasprotovali uradnim Covid ukrepom [12]. Intuitivno bi se nam zdelo, da so to skupine, ki ne “zaupajo znanosti”, avtorji uporabijo termin “anti-znanost” 2 Politična znanost (anti-science), obstaja tudi variacija “zanikalec znanosti” Zagovarjam stališče, da bi “zaupanje v znanost” moralo pomeniti (science-denier). Vendar se je izkazalo, da ti ljudje niso klasični zaupanje v znanost kot proces in metodo, ne pa v njen človeški oziroma stereotipni zanikalci znanosti, v resnici sploh ne element (znanstveniki in institucije), ki je dovzeten za razne nasprotujejo znanosti kot taki in da so nadpovprečno znanstveno pristranosti in konflikte interesa, zaradi katerih trpi pismeni. Nasprotovali so uradni (politično podprti) znanosti, verodostojnost znanstvenih zaključkov. Znanstvenega procesa v razlikovanju med uradno in neuradno znanostjo ter praksi seveda ni brez človeškega elementa, ki ta proces izvaja, avtoritarnemu odnosu stroke do laikov. Zagovarjali so torej vendar človeški element v tej izvedbi tudi ni nezmotljiv. Zato egalitarno znanost, kjer ima vsakdo dostop do podatkov in velja zaupati v process, v človeški element pa ne povsem. možnost oblikovanja svojih zaključkov [12]. Posledično moramo ugotoviti, ali se uporaba tega slogana v Avtorji študije se s tem niso strinjali in so trdili, da Covid-zahodni družbi sklada s tovrstnim razumevanjem ali ne. V skeptiki “spodkopavajo uradne znanosti s spretno manipulacijo kolikor se ne, in za tem stoji pričakovanje slepega zaupanja podatkov”. Ta trditev se mi zdi bizarna – kako lahko želja po znanstvenikom in institucijam, je to znak dogmatizma in intersubjektivnem preverjanju s strani visoko znanstveno institucionalizacje znanosti, ki sta močno povezani s politizacijo. pismenih posameznikov, ki želijo nepristransko ovrednotiti Carl Schmit je znan po svoji definiciji politike kot presojanju podatke 1 “spodkopava uradno znanost”? Ni to kvečjemu na podlagi dihotomije prijatelj/sovražnik, pri čemer je prijatelj koristno, saj je po Popperju 2 ravno falsifikacija gonilo nekdo s komer si delim interese, sovražnikovim interesom pa znanstvenega napredka, ki je v času pandemije še toliko bolj nasprotujem [18]. Politično vrednotenje dogajanj in dejanj torej ključen? Rekel bi, da načeloma je, vendar ne v ne temelji na splošnih načelih, temveč poteka na podlagi institucionalizirani znanosti, kjer akademske institucije želijo identitete udeleženih subjektov in uporabnosti njegovih posledic obdržati monopol nad produkcijo znanja. Institucionalizacijo 1 Lee et al (2021): “These users want to understand and analyze the information for 2 Popperjev model falsifikacije ima sicer svoje težave, vsekakor pa gonilo themselves, free from biased, external intervention.”, str. 12 znanstvenega napredka ni izogibanje možnostim falsifikacije. 23 sicer lahko razumemo kot mehko obliko politizacije, vendar to postavil teorijo kognitivne disonance – ljudje se držimo očitno še ni indikator politizacije v pravem pomenu besede. neresničnih stališč, ker težimo k ujemanju stališč, vedenja in Žal pa je med Covid krizo prišlo tudi do slednje. V samopodobe [6]. Za naštete fenomene predlagam nadpomenko institucionalizirani znanosti vlada kredencializem – merilo ideje socialnega presojanja in sklepanja5 – presojanja in sklepanja v je znanstveni in akademski prestiž znanstvenika, ki jo predlaga. skladu s svojo skupinsko identiteto, konsenzom ingrupe ali Vendar med Covid krizo niti znanstveni prestiž avtorja ni bil stališčem ingrupne intelektualne avtoritete, kar pogosto vodi do zadosten pogoj za sprejemanje neke ideje. Tako se je npr. dr. fenomena, ki ga Perkins (po navedbi Barona) poimenuje myside Robert Malone moral soočiti z deplatformiranjem zaradi bias [1]. Socialnega presojanja se po mojem mnenju “širjenja dezinformacij” – Twitter mu je deaktiviral račun [15] poslužujemo na vseh družbeno-relevantnih področjih, kjer po nastopu na Roganovem podcastu, kjer je izrazil nestrinjanje z nimamo motivacije, sposobnosti ali predznanja za sistematično uradnim konsenzom glede Covida in zajezitvenih ukrepov, ter oblikovanje lastnega stališča. svoje stališče znanstveno argumentiral 3 . Malone je sicer V to kategorijo zaradi svoje kompleksnosti spada večina znanstvenik – mednarodnega renomeja 4 – vendar očitno ni znanstvenih tem, še posebej tistih, ki so družbeno oziroma izpolnjeval kriterijev za “zaupanje znanosti”. Kaj je torej znanost, politično relevantne, vključno s pandemijo Covida-19 in z njo na katero se je med Covid krizo nanašal slogan “zaupajmo povezanimi ukrepi. Pinker govori o fokusnih točkah, javno znanosti”? Ugotovili smo, da se ne nanaša na proces vidnih in relevantnih dogodkih in dogajanjih, ki jih vidi znanstvenega raziskovanja in niti na individualne znanstvenike z posameznik in se hkrati zaveda, da so vidni tudi drugim dovoljšno mero prestiža. Moja teza je torej, da se je beseda prebivalcem družbe [17]. Fokusne točke, oziroma spektakli, “znanost” nanašala na uradno, torej politično-podprto znanost pogosto postanejo politizirane – to so močno družbeno oziroma znanost režima. Tukaj se lahko navežem na Foucaltov relevantna dogajanja, do katerih se je potrebno opredeliti. Že režim resnice, kjer je ideja resnice politično in ideološko sama potreba po opredelitvi je političnega značaja, ker ne umeščena – diskurz in metode produkcije resnice so omejeni, dopušča nevtralnosti, zgolj izbiro enega izmed dveh polov. Ko je hkrati obstaja skupina ljudi, ki ima monopol nad razglašanjem prisotna binarna polarizacija, pa je prisotna tudi dihotomija družbene resnice [8]. prijatelja (podpornika uradnih ukrepov) in sovražnika Med Covid krizo so vlogo razsodnika resnice prevzeli (nasprotnika uradnih ukrepov). Fokusne točke torej aktivirajo in znanstveniki režima – uradni Covid komentatorji (kot npr. dr. okrepijo vrojeno tendenco človeka po socialnem presojanju, v Fauci v Ameriki, dr. Krek in dr. Beović v Sloveniji), vlogo tem primeru o vsebini same fokusne točke. Ko je zaupanje “čuvaja” resnice pa mediji in socialni mediji, ki so tako ali znanosti postalo fokusna točka, kar se je zgodilo med Covid krizo drugače utišali znanstvenike, ki so želeli izraziti kakršno koli (če ne še prej), se je torej navzelo političnih konotacij in postalo nestrinjanje z uradnim konsenzom. Covid krizo je torej označevalec za prijatelje in sovražnike režima – definirane kot zaznamovala močna politizacija znanosti, saj je pravico do zaupnike in zanikalce znanosti (včasih teoretike zarote). širjenja (znanstvenih) resnic nudila predvsem podpora Zaupanje v znanost je torej družbenopolitični problem. Stran, na (prijateljstvo) režima, ki se je odražala v podpori uradnega kateri se nekdo nahaja, je prej merilo politične opredeljenosti kot konsenza glede spopadanja s pandemijo. Posledično trdim, da samega zaupanja v znanost v klasičnem pomenu izraza, ali odraz slogan “zaupajmo znanosti” ni predstavljal klica k epistemski globljih filozofskih načel. Drugače povedano, izražanje racionalnosti in sistematičnemu presojanju znanstvenih izjav, (ne)zaupanja v znanost v kakršnem koli socialnem kontekstu je temveč ravno nasprotno – emocionalno in politično prežet sklic politična uniforma, zato je to prej signal privrženosti ustaljeni na avtoriteto. Cilj je bil sprejemanje stališč intelektualnih politiki kot pokazatelj odnosa do raziskovalne dejavnosti, ki ji avtoritet režima, ne pa samostojni razmislek. pravimo znanost. 3 Socialno presojanje 4 Politični in spoznavni razhod Tematika letošnje konference je “kognitivni vidiki zaupanja v V obdobju politične polarizacije zaradi socialnega presojanja in znanost”. Moj cilj je pokazati, da je zaradi politizacije znanosti politizacije znanosti pogosto pride do spoznavnega razhoda – na in zaupanja v znanost večina teh kognitivnih vidikov pod eni strani imamo množico ljudi, ki takorekoč zaupa znanosti vplivom socialnih pritiskov. oziroma uradnim virom in zgodbam, na drugi pa množico ljudi, V socialni psihologiji obstaja veliko raziskav in teorij na temo ki “zanika znanost” – torej zavrača uradne vire in zgodbe, ter oblikovanja in spremembe stališč ter presojanja novih informacij. oblikuje svoja stališča s pomočjo alternativnih virov. Giner-Sorolila in Chaiken sta poimenovala koncept motiviranega Pojavita se vsaj dve različni “socialni resničnosti”, dve sklepanja, kjer sistematično sklepamo z namenom potrditi točno različni interpretaciji vsebine fokusne točke. Imamo torej ljudi, določeno stališče [9]. Caccioppo in Petty sta postavila ki v grobem sprejemajo uradno zgodbo in ljudi, ki jo v grobem dvoprocesni model spremembe stališč, kjer centralno zavračajo (seveda pa sta to sprejemanje in zavračanje procesiranje upošteva predvsem vsebino sporočila, periferno pa kontinuum), v primeru Covida se to nanaša na stališča do mask, lastnosti sporočevalca in socialni kontekst [2]. Festinger pa je cepljenja in drugih uradnih ukrepov. To je v veliki meri posledica 3 https://open.spotify.com/episode/3SCsueX2bZdbEzRtKOCEyT 5 Ta koncept sem podrobneje razdelal v članku Social Reasoning and the 4 Malone na https://www.rwmalonemd.com: “I am an internationally recognized Politicization of Science During the Covid Pandemic, ki bo objavljen Decembra v scientist/physician and the original inventor of mRNA vaccination as a technology-reviji Mankind Quarterly [16]. I have approximately 100 scientific publications with over 12,000 citations of my work (per Google Scholar with an “outstanding” impact factor rating committees).” 24 razlik v zaznavanju zaupanja vrednih oziroma verodostojnih Odgovor je, seveda, režim in “znanost” režima. Vidimo torej, virov v obeh (ali vseh) skupinah ljudi. Vir, ki je verodostojen za da je spoznavni razhod med podporniki uradne narative in eno skupino nikakor ni verodostojen za drugo, to presojanje o kontranarative posledica aktivno ustvarjenega političnega verodostojnosti pa je politične narave. Torej, spoznavni razhod razhoda s strani režima in njegovih ideoloških aparatov, ki v je posledica političnega razhoda, ne obratno. Oziroma, kot bi interakciji z javnostjo ustvarjajo koncept zaupanja znanosti, rekel Foucault, znanje izvira iz moči. In šele nato spoznavni zanikanja znanosti in teorij zarote. Akademiki in drugi razhod perpetuira političnega – sprejemanje ene ali druge raziskovalci imamo edinstveno možnost izpostavljanja napak interpretacije (označeno kot zaupanje znanosti ali teoriziranje režima, ampak lahko to dosežemo zgolj, če znanost zaščitimo zarote) je politična uniforma, ki signalizira pripadnost enemu od pred politizacijo. Prvi korak k depolitizaciji znanosti pa je po političnih polov. mojem mnenju prepoznavanje koncepta zaupanja v znanost kot Kljub temu pa pomanjkljivo znanje, do katerega pride v politične uniforme in posledično zavračanje vseh dihotomij, ki primeru cenzure nasprotujočih stališč, nosi svoje posledice – jih ustvari. pogosto negativne. V zadnjih mesecih prihaja vedno več raziskav in medijskih objav, ki izpostavljajo destruktivne posledice LITERATURA določenih Covid ukrepov – ekonomska škoda, ki so jo povzročili [1] Baron, J. (2000), Thinking and deciding (3rd ed.), New York: Cambridge lockdowni [20], zaviranje razvoja otrok zaradi obveznega University Press [2] Cacioppo, J.T. in Petty, R.E. (1984). The Elaboration Likelihood Model nošenja mask [23] in njihova splošna neučinkovitost [19], of Persuasion. NA - Advances in Consumer Research, 11, 673-675. neučinkovitost cepiv pri zaščiti pred okužbo s Covidom [5] in [3] CDC (2022). CDC streamlines COVID-19 guidance to help the public možnost nevarnih stranskih učinkov pri določenih demografskih better protect themselves and understand their risk. Sneto iz https://www.cdc.gov/media/releases/2022/p0811-covid-guidance.html skupinah, npr. nosečnicah [4]. Ameriški CDC je sicer pred [4] Clark County Today (2022). FDA knew huge percentage of women in kratkim spremenil svoje smernice za spopadnje s Covidom – zdaj Pfizer trial suffered miscarriages. Sneto iz: https://www.clarkcountytoday.com/news/fda-knew-huge-percentage-of-so enake za cepljene in necepljene posameznike, kar implicira women-in-pfizer-trial-suffered-miscarriages/ enako stopnjo tveganosti obeh skupin [3]. Vendar se moramo [5] Comber, J. in Madhava, S. (2022). After Data Show Vaccinated at Higher vprašati, zakaj šele zdaj? Različni ljudje in institucije po svetu so Risk of Dying from COVID, Canadian Province Ends Monthly Reports. Global Research. Sneto iz: https://www.globalresearch.ca/after-data-tako ali drugače opozarjali na morebitne negativne posledice show-vaccinated-higher-risk-dying-from-covid-canadian-province-ends-uradnih Covid ukrepov, vendar so bili tako ali drugače utišani. monthly-reports/5790795 [6] Festinger, L. (2022). A Theory of Cognitive Dissonance (Anniversary ed.). Tukaj torej vidimo, da imata politizacija znanosti in dogmatični Stanford University Press. odnos do tako-imenovanega “strokovnega konsenza” v naši [7] Feynman, R. (n.d.). What is Science? Sneto iz: http://www.feynman.com/science/what-is-science/ poznanstvenjeni družbi obsežne negativne posledice. [8] Fontana A. & Pasquino, P. (n.d.) Truth and Power (interview with V svetu, kjer se zdi, da lahko motiviran laik z dovoljšno mero Foucault). Sneto iz znanstvene pismenosti v enem tednu iskanja člankov na Google https://www2.southeastern.edu/Academics/Faculty/jbell/foucaulttruthpo wer.pdf Scholar doseže osnovno razumevanje (ali vsaj aproksimacijo le- [9] Giner-Sorolila, R., & Chaiken, S. (1997). Selective Use of Heuristic and tega) nekega znanstvenega področja, uradne znanstvene Systematic Processing Under Defense Motivation. Personality and Social Psychology Bulletin, 23(1), 84–97. doi:10.1177/0146167297231009 institucije niso več edini možni vir znanja. In v skladu s tem se [10] Hacking, I. (1995). The looping effects of human kinds. In D. Sperber, D. moramo tudi ravnati in priznavati veljavnost izvenkonsenzualnih Premack, & A. J. Premack (Eds.), Causal cognition: A multidisciplinary stališč, v kolikor so podprta z argumenti in dokazi. debate (pp. 351–394). Clarendon Press/Oxford University Press. [11] Jigsaw (2021). Psychological Inoculation: New Techniques for Fighting Na žalost pa Googlov think tank Jigsaw in Svet za Online Extremism. Medium. Sneto iz: družboslovno raziskovanje (Social Science Research Council, https://medium.com/jigsaw/psychological-inoculation-new-techniques-for-fighting-online-extremism-b156e439af23 SSRC), tako kot Lee in kolegi povlečeta ravno obraten zaključek. [12] Lee, C., Yang, T., Inchoco, G. D., Jones, G. M., & Satyanarayan, A. Laikom ne želita prepustiti, da si sami ustvarijo stališče in sami (2021). Viral Visualizations: How Coronavirus Skeptics Use Orthodox presojajo med informacijami in dezinformacijami, oziroma med Data Practices to Promote Unorthodox Science Online. Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems. znanjem in lažnimi novicami. Nasprotno, Jigsaw predstavlja https://doi.org/10.1145/3411764.3445211 koncept “pre-bunkinga” oziroma psihološke inokulacije, [13] Lévi-Strauss, C. (1987). Introduction to Marcel Mauss. Routledge. [14] Lyotard, J.-F. (1984). The Postmodern Condition: A Report on vnaprejšnjega zavračanja možnih heterodoksnih stališč v obliki Knowledge. University of Minnesota Press. kratkih sporočil, ki predstavijo protiargumente in poslušalcu [15] Malone, R. (2021). Permanently suspended on Twitter. Who is Robert Malone. Sneto iz: https://rwmalonemd.substack.com/p/permanently- olajšajo zavračanje tega stališča v prihodnosti [11]. SSRC pa suspended-on-twitter skuša ugotoviti kako maksimizirati povpraševanje po Covid [16] Marušič, J. Ž. (2022). Social Reasoning and the Politicization of Science cepivih – tako da dijake in študente nauči prepoznavati During the Covid Pandemic. Mankind Quarterly. Sprejeto v objavo. [17] Pinker, S. (2021). Rationality: What Is It, Why It Seems Scarce, Why It “dezinformacije o cepivih”. sporočevalce opremi z ustreznimi Matters. Viking Press. “sporazumevalnimi strategijami” in na družbenih omrežjih [18] Schmitt, C. (2007) The Concept of the Political. University of Chicago Press. oblikuje “(demografsko in geografsko) prilagojena sporočila” [19] Spira B (April 19, 2022) Correlation Between Mask Compliance and [21]. COVID-19 Outcomes in Europe. Cureus 14(4): e24268. Spet se moramo vprašati, kdo razlikuje med informacijo in doi:10.7759/cureus.24268 [20] Sumption, J. (2022). Little by little the truth of lockdown is being dezinformacijo, med ortodoksnimi in heterodoksnimi stališči. Je admitted: it was a disaster. The Times. Sneto iz: to znanost, politika ali politizirana znanost? In nadalje, ne bi to https://www.thetimes.co.uk/article/little-by-little-the-truth-of-lockdown- is-being-admitted-it-was-a-disaster-5b5lrlgwk morala biti pravica in dolžnost vsakega odraslega državljana v [21] The Rockefeller Foundation. Mercury Project to Boost Covid-19 demokratični in egalitarni državi? Če si posameznik ne more, Vaccination Rates and Counter Public Health Mis- and Disinformation in 17 Countries Worldwide. Sneto iz: oziroma ne sme sam ustvariti mnenja, čemu potem služi https://www.rockefellerfoundation.org/news/mercury-project-to-boost-demokracija? 25 covid-19-vaccination-rates-and-counter-public-health-mis-and- disinformation-in-17-countries-worldwide/ [22] Ule, A. (2006). Znanost, družba, vrednote. Aristej. [23] Watson, S. (2022). CNN Medical Analyst Who Fiercely Advocated Masking Now Admits It ‘Harmed’ Her Own Son’s Development. Summit News. Sneto iz: https://summit.news/2022/08/24/cnn-medical-analyst-who-fiercely-advocated-masking-now-admits-it-harmed-her-own-sons-development/ 26 Filozofski in psihološki vidiki človeške racionalnosti Philosophical and psychological aspects of human rationality Nastja Tomat Oddelek za filozofijo Filozofska fakulteta UL Ljubljana, Slovenija nastja.tomat@ff.uni-lj.si POVZETEK to the shift from the idealized view of human rationality to a more moderate one. In addition to research on rational action, study of Človeška racionalnost je kompleksen pojem, ki se nanaša na rational beliefs is another field of inquiry that connects široko paleto našega spoznavanja in delovanja. Obstajajo investigation of rationality with fundamental questions in številne opredelitve racionalnosti; Ronald de Sousa razlikuje epistemology. med kategorično in normativno racionalnostjo, govorimo lahko o instrumentalni ali široki racionalnosti ali o racionalnosti kot KEYWORDS logičnem sklepanju. Vprašanja o racionalnosti so tesno bounded rationality, ecological rationality, rationality of belief, prepletena s preučevanjem odločanja. Normativne teorije odločanja racionalno vedenje opredelijo kot tisto, ki vodi do izida heuristics and biases z največjo pričakovano koristnostjo, deskriptivne teorije pa preučujejo, kako se odločanje v vsakdanjem življenju dejansko 1 UVOD poteka. K odmiku od idealiziranega pogleda na racionalnost so pripomogli program hevristik in pristranosti, ki sta ga osnovala Vprašanje, ali smo ljudje racionalna bitja, še zdaleč ni enostavno. Daniel Kahneman in Amos Tversky, koncept omejene Odgovor se že stoletja izmika znanstvenikom različnih disciplin racionalnosti, ki ga je predstavil Herbert A. Simon, ter delo od ekonomije in psihologije do filozofije in kognitivne znanosti. Gerda Gigerenzerja in sodelavcev, ki preučujejo ekološko Človeška racionalnost je tema, ki se je lahko lotevamo iz racionalnost. Poleg racionalnosti dejanj lahko govorimo tudi o številnih vidikov in z uporabo različnih metod, zato ni racionalnosti prepričanj, kar preučevanje racionalnosti poveže s nenavadno, da danes na tem področju obstaja ogromno polje temeljnimi vprašanji s področja epistemologije. razprav in raziskav. V veliki razpravi o racionalnosti, kot so to poimenovali v kognitivni znanosti, obstajata dva nasprotujoča si KLJUČNE BESEDE pogleda. Na enem polu so avtorji, ki zagovarjajo, da so človeško sklepanje, presojanje in odločanje, ki so del racionalnega omejena racionalnost, ekološka racionalnost, racionalnost vedenja, polni pomanjkljivosti in pristranosti ter da jih je mogoče prepričanj, hevristike in pristranosti izboljšati; zagovorniki takšnega pogleda v veliki meri izhajajo iz ABSTRACT programa hevristik in pristranosti, ki sta ga osnovala psihologa Daniel Kahneman in Amos Tversky. Raziskovalci na drugem Human rationality is a complex topic that encompasses a wide polu pa takšnemu pogledu na racionalnost nasprotujejo in trdijo, range of cognitive processes and behavior. Many definitions of da so kriteriji normativnih teorij racionalnosti neustrezni ter da rationality exist, one of them being Ronald de Sousa's notion of izsledki empiričnih raziskav, ki pričajo o sistematičnih odklonih categorical and normative rationality. Some authors distinguish od omenjenih kriterijev, še ne zadostujejo za sklep, da smo ljudje between instrumental and broad conception of rationality, while iracionalni [1, 2, 3]. others define rationality in terms of logical reasoning. The study Namen prispevka je podati pregled izbranih pogledov na of rationality is intertwined with research in the field of decision človeško racionalnost. Začela bom z definicijo filozofa Ronalda making. Normative theories define rationality as behavior that de Sousa, nadaljevala pa z dvema opredelitvama racionalnosti, leads to the outcome with the greatest expected utility, while med katerima se v literaturi pogosto razlikuje: instrumentalno in descriptive theories examine how people actually make decisions široko. Na primeru Wasonove naloge izbire kart – ene najbolj in everyday life. Kahneman and Tversky's heuristics and biases uporabljenih nalog pri empiričnem preučevanju sklepanja – bom program, Herbert A. Simon's concept of bounded rationality and opisala pogled, ki racionalnost povezuje z logičnim sklepanjem Gerd Gigerenzer's study of ecological rationality all contributed ter je še vedno vpliven zlasti na področju filozofije. Poleg logike je področje, ki je prav tako prepleteno s preučevanjem Permission to make digital or hard copies of part or all of this work for personal or racionalnosti, odločanje. Opisala bom, kakšno sliko racionalnosti classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full prikazujejo normativne teorije odločanja ter kako se je kot kritika citation on the first page. Copyrights for third-party components of this work must takšnega pogleda izoblikoval program hevristik in pristranosti, ki be honored. For all other uses, contact the owner/author(s). Information Society 2022, 10–14 October 2022, Ljubljana, Slovenia je še danes eden najvplivnejših okvirjev za preučevanje © 2022 Copyright held by the owner/author(s). odločanja in presojanja. Nato bom predstavila koncept omejene racionalnosti, ki ga je oblikoval Herbert A. Simon in je 27 pomembno vplival na razumevanje in pojmovanje racionalnosti, postavlja. Prednost takšnega pristopa je v tem, da lahko ter koncept ekološke racionalnosti, ki ga preučujejo Gerd postavimo norme, ki služijo kot kriterij racionalnosti, ter Gigerenzer in sodelavci ter se naslanja na Simonovo delo. V spremljamo, v kolikšni meri in pod kakšnimi pogoji ljudje od njih zadnjem delu se bom odmaknila od empiričnih raziskav odstopamo. Po drugi strani pa se zdi preučevanje racionalnosti le odločanja in presojanja ter opisala nekatera vprašanja, ki jih iz instrumentalnega vidika preozko – če se osredotočamo samo odpira raziskovanje racionalnosti prepričanj – teme na presečišču na ciljno usmerjeno vedenje, izpustimo pa vprašanja o ciljih, preučevanja racionalnosti in epistemologije. normah in vrednotah, zanemarimo velik in pomemben del človeškega delovanja [1]. John Searle v svoji knjigi o racionalnosti navaja primer znanih raziskav o inteligentnosti 2 DE SOUSOVA OPREDELITEV opic, ki jih je psiholog Wolfgang Köhler izvajal na Tenerifih. V RACIONALNOSTI eksperimentih se je izkazalo, da so opice sposobne reševanja Filozof Ronald de Sousa najprej razlikuje med kategorično in problemov z vpogledom; da bi dosegle na strop obešene banane, normativno racionalnostjo. Pri kategorični racionalnosti je do katerih niso mogle priti s skakanjem, so uporabile škatle in nasprotje racionalnega aracionalno vedenje. Racionalno je palico [6]. Iz intrumentalnega vidika so se opice torej vedle takšno vedenje, ki ga vodijo določeni razlogi, aracionalno pa racionalno in Searle meni, da tudi racionalnost človeka še vedno takšno, ki ga ne vodi mišljenje ali izbira. Pri kamnu, ki ga vržemo presojamo na podoben način. V klasičnih modelih racionalnosti skozi okno, ali človeku, ki se spotakne in pade v grm kopriv, ne je človeška racionalnost pravzaprav le kompleksnejša verzija govorimo o (i)racionalnosti – pri prvem gre namreč za pojav, ki šimpanzje. Searle v nadaljevanju opozarja na pomanjkljivosti uboga zakone fizike, pri drugem pa za dejanje, ki ga ni vodila takšnega pojmovanja racionalnosti in opozarja na pomembnost izbira. Pri normativni racionalnosti pa razlikujemo med ločevanja vedenja na podlagi želja in na podlagi razlogov [7]. racionalnim in iracionalnim vedenjem. Racionalno vedenje je V odgovor na pomanjkljivosti instrumentalnega pristopa so tisto, ki je ustrezno utemeljeno z določenimi razlogi, normami ali se pojavila širša pojmovanja racionalnosti, ki upoštevajo tudi vrednotami, iracionalno pa tisto, ki se od temu pogoju na tak ali cilje, prepričanja, norme in vrednote, ki usmerjajo naše vedenje. drugačen način ne zadostuje. De Sousa pravi, da lahko o ljudeh Te teorije se med drugim ukvarjajo z vprašanji o racionalnosti kot o racionalnih živalih govorimo samo, če sprejmemo, da smo samih ciljev [5] ter o vedenju, ki nima samo instrumentalne ljudje racionalni v kategoričnem smislu in kot taki tudi sposobni funkcije [1]. Filozof Robert Nozick na primer govori o konceptu iracionalnega vedenja [4]. simbolne koristnosti in pravi, da imajo naša dejanja neodvisno Če kategorične racionalnosti ne pripisujemo dogodkom, ki jih od instrumentalne tudi simbolno vrednost, ki bi jo morale lahko zadostno razložimo z naravnimi zakoni, ali to pomeni, da vključevati vse formalne teorije racionalnosti in odločanja. Ker z njimi ne moremo razložiti človeškega vedenja? Zmernejša živimo v socialno in simbolno kompleksnem okolju, naša dejanja interpretacija pravi, da je človeško vedenje podvrženo naravnim služijo tudi namenom, ki presegajo doseganje ozko zastavljenih zakonom, vendar ti ne ponujajo zadostne razlage. Kot primer de ciljev, na primer temu, da sebi in drugim sporočamo, kakšne Sousa navaja igro šaha, ki ga moramo razložiti s pravili igre – in osebe smo [8]. Podobno ekonomist Shaun H. Heap kot protipol ta niso naravni zakoni. Močnejša interpretacija pa pravi, da instrumentalni racionalnosti postavlja ekspresivno racionalnost. vedenje racionalnih bitij, vključno s človekom, na nek način Ko izvajamo dejanja, ki so ekspresivno racionalna, presega zakone narave. De Sousa meni, da je tako stališče opredeljujemo in raziskujemo lastna prepričanja in vrednote. Ne absurdno, saj bi predpostavljalo čudež ali pa vsaj to, da zakonov gre torej za enosmerno povezavo med vrednotami in delovanjem, narave ne razumemo pravilno. Zagovarja, da moramo človeka temveč za povratno zanko, kjer z dejanji vrednote tudi obravnavati kot bitje, ki je kot vsa ostala podvržen zakonom konstruiramo, spremljamo in prilagajamo [9]. narave; razliko med človekom in ostalimi bitji je potrebno iskati v zakonih narave in ne v lastnostih, ki bi le-te na nek način presegale. Če privzamemo, da se racionalnost nanaša na misli in 4 RACIONALNOST IN LOGIČNO dejanja, lahko razlikujemo med dvema ključnima spremembama MIŠLJENJE tako na nivoju evolucije kot razvoja posameznika: prva je razvoj Najbrž eden od najstarejših kriterijev racionalnosti je sledenje od golega zaznavanja objektov do zmožnosti tvorbe pravilom logičnega sklepanja in verjetnostnega računa [1]. Ena reprezentacij, druga pa razvoj od avtomatskih vedenjskih od najbolj preučevanih nalog, ki se uporablja v empiričnih odzivov do zmožnosti oblikovanja namer ter želja ter vedenja na raziskavah sklepanja, je Wasonova naloga izbire kart [10, 11], ki podlagi le-teh [4]. ima naslednjo obliko: »Na mizi so štiri karte. Vsaka ima na eni strani številko, na drugi pa barvo. Katere karte je potrebno obrniti, da testiraš pravilo: če je na eni strani sodo število, je na 3 INSTRUMENTALNA IN ŠIROKA drugi strani rdeča barva?« RACIONALNOST V literaturi se pogosto pojavlja razlikovanje med ožjim, instrumentalnim in širokim pojmovanjem racionalnosti [1, 4]. Instrumentalno racionalnost opredelimo kot vedenje, ki nas približa doseganju zastavljenega cilja glede na mentalne in fizične vire, ki so nam na voljo. Povedano drugače, racionalno je tisto vedenje, ki optimizira doseganje ciljev, pri čemer se ne ukvarjamo s tem, kakšni ti cilji so in kako si jih posameznik 28 Slika 1: Primer Wasonove naloge izbire kart. primeru, ko posameznik preferira alternativo A pred B in B pred C, preferira tudi A pred C. Če aksiomi držijo, lahko vsaki V zgoraj navedenem primeru je pravilni odgovor, da je alternativi pripišemo določeno koristnost in racionalno odločanje potrebno obrniti karto s številko 8, s čimer preverimo modus je tisto, ki privede do izida z najvišjo koristnostjo [22]. ponens, in karto rjave barve, s čimer preverimo modus tollens. Normativne teorije pred odločevalce torej postavljajo stroge Večina udeležencev pri takšni nalogi poda odgovor, da je zahteve in kmalu so se začela pojavljati vprašanja, če se ljudje v potrebno obrniti karto s številko 8 in karto z rdečo barvo, vendar vsakdanjem življenju resnično odločamo na tak način. gre pri slednjem za napako zatrjenega konsekvensa. V več kot Kahneman in Tversky sta leta 1979 objavila članek, v katerem petdesetih letih od izvirne objave je bila naloga uporabljena v sta pokazala, da ljudje sistematično kršimo aksiome racionalnosti, ogromnem številu raziskav, kjer so avtorji manipulirali z na katerih slonijo normativne teorije. Svoje ugotovitve sta strnila različnimi spremenljivkami, ki bi lahko vplivale na izvedbo v teorijo obetov, ki nadgrajuje teorijo pričakovane koristnosti in naloge, še danes pa ni enotne razlage za majhen delež pravilnih razlaga, kako se ljudje odločamo pod pogojem tveganja [23]. rešitev; ena od interpretacij je, da se večinoma osredotočamo na Kahneman in Tversky sta dolga leta preučevala presojanje in potrjevanje hipoteze, manj pa preverjanje pogojev, ki bi hipotezo odločanje in osnovala raziskovalni okvir, ki ga poznamo pod ovrgli [12, 13, 14]. Še ena ugotovitev je, da so udeleženci imenom »program hevristik in pristranosti«. V številnih pogosto nagnjeni k izbiri kart, ki so eksplicitno omenjene v raziskavah sta pokazala, da ljudje v negotovih pogojih pogosto navodilu [15]. Eno od opažanj je, da se delež pravilnih rešitev uporabljamo hevristike – miselne bližnjice, ki olajšujejo poveča, če namesto abstraktnih uporabimo konkretne primere, reševanje problemov, so hitre, varčne in zahtevajo manj napora kar nakazuje na to, da se pri logičnem sklepanju oz. tesitranju – kar vodi do sistematičnih napak v presojanju in odločanju, ki hipotez ne zanašamo le na obliko argumentov, temveč tudi na sta jih poimenovala kognitivne pristranosti. Ljudje pogosto ne vsebino [16]. Delež pravilnih rešitev je še večji, če uporabimo upoštevamo pravil logike in verjetnostnega računa, smo slabi deontična pravila. Če morajo udeleženci na primer preverjati intuitivni statistiki, zaključujemo brez ustreznih dokazov, slabo pravilo »Če piješ alkohol, moraš biti starejši od 18 let«, na mizi napovedujemo lastne prihodnje preference in smo podvrženi pa imajo karte s številkami 16 in 25 ter z napisi »pivo« in številnim dejavnikom, ki na tak ali drugačen način »kokakola«, večina pravilno izbere karti s številko 16 in napisom »neupravičeno« vplivajo na naše presoje. Doprinos programa »pivo«. Testiranje hipotez nam gre očitno torej bolje, ko moramo hevristik in pristranosti je ravno v poudarjanju tega, da ljudje preverjati morebitne kršitve socialnih pravil [17]. Ena od nismo racionalni v okviru normativnih teorij, temveč da interpretacij, ki temelji na evolucijski psihologiji, je, da sklepanje odločanje in presojanje v vsakdanjem življenju potekata drugače ni le splošen, od vsebine neodvisen proces, temveč je v ozadju [24, 25, 26]. več specializiranih procesov, eden izmed katerih je namenjen Delo Kahnemana in Tverskyja je bilo skozi leta deležno reševanju problemov v kontekstu socialnih izmenjav in kršitev različnih kritik. Če smo ljudje resnično tako podvrženi socialnih pogodb [17]; ta interpretacija je deležna številnih kritik sistematičnim napakam v presojanju in odločanju, kako je sploh [18]. Nekateri avtorji pa menijo, da je logično pravilna rešitev mogoče, da se dovolj učinkovito odzivamo na okolje, da Wasonove naloge v konfliktu z načinom, kako v vsakdanjem preživimo? Različni avtorji so ponudili alternativne življenju testiramo hipoteze, ter zagovarjajo, da je način, kako se interpretacije izsledkov, ki naj bi izražali pristranosti v mišljenju. udeleženci lotijo reševanja, v resničnem življenju adaptiven. Po Ena vrsta interpretacij se ukvarja z razlago odgovorov na naloge njihovem neuspešno reševanje naloge torej ne služi kot dokaz znotraj laboratorijskih pogojev, druga pa z vprašanjem, kaj nam iracionalnosti [19, 20]. To se sklada s pogledom, da logičnega ti odgovori povedo o sklepanju, presojanju, odločanju in mišljenja ne gre vedno in apriori enačiti z racionalnostjo, temveč reševanju problemov v vsakdanjem življenju. Že znotraj je ustreznost takšnega mišljenja odvisna tudi od konteksta [1]. laboratorijskega konteksta ni vedno enoznačno, ali je določen odgovor na nalogo pravilen ali napačen. Primer tega so različne interpretacije že omenjene Wasonove naloge izbire kart. 5 RACIONALNOST IN ODLOČANJE Oaksford in Chater na primer menita, da naloga ne ocenjuje Pojem racionalnosti je tesno prepleten s preučevanjem odločanja deduktivnega sklepanja, temveč verjetnostno. Če privzamemo, in presojanja. Znotraj vedenjskega preučevanja odločanja ločimo da kriterij za pravilne odgovore ni upoštevanje pravila med normativnimi, deskriptivnimi in preskriptivnimi pristopi. falsifikacije, temveč izbira najbolj informativnih kart v skladu s Normativne teorije se osredotočajo na to, kako bi se ljudje morali teorijo optimalne selekcije podatkov (ang. optimal data odločati, da bi prišli do izida, ki ima zanje največjo koristnost, selection), lahko nekatere odgovore udeležecev smatramo kot deskriptivne teorije preučujejo, kako človeško odločanje v pravilne, tudi če ne sledijo pravilom formalne logike [19, 20]. resničnem življenju dejansko poteka, preskriptivne pa želijo Zagovorniki druge vrste interpretacij pa segajo izven laboratorija zmanjšati vrzel med prvima dvema in osnovati predloge za in menijo, da so »napačni« odgovori udeležencev iz izboljšanje odločanja [21]. evolucijskega, adaptivnega vidika pravzaprav smiselni. Prevladujoč model normativnega odločanja pod pogojem Odgovori, ki jih v umetno ustvarjenih problemih v tveganja je bila dolgo časa teorija pričakovane koristnosti, ki sta laboratorijskem eksperimentiranju razlagamo kot napake, imajo jo v knjigi Theory of Games and Economic Behaviour leta 1944 v vsakdanjem življenju prilagoditveno vlogo in zato morda ni predstavila John von Neumann in Oskar Morgenstern. Teorija upravičeno, da jih jemljemo kot dokaz človeške iracionalnosti temelji na aksiomih, ki se nanašajo na odločevalčeve preference. [27, 28, 29]. Med drugim predpostavljajo, da ima posameznik popoln, urejen in tranzitiven nabor preferenc; to pomeni, da lahko za vsak par alternativ določi, v kakšnem odnosu sta, velja pa tudi, da v 29 6 OMEJENA RACIONALNOST ostri kritiki programa hevristik in pristranosti. Zagovarjajo, da so hevristike lahko učinkovita orodja mišljenja in da poseganje po Še en koncept, ki je pomembno vplival na odmik od njih v nekaterih situacijah, sploh takšnih z visoko stopnjo idealiziranih, normativnih teorij odločanja, je bil koncept negotovosti, lahko pojmujemo kot racionalno. Ukvarjajo se s omejene racionalnosti, ki ga je v 50. letih prejšnjega stoletja tako imenovano ekološko racionalnostjo, kjer je poglavitno predstavil Herbert A. Simon. Simon je menil, da je pojem vprašanje, katera strategija v določeni situaciji vodi do boljših globalne racionalnosti, ki naj bi jo posedoval človek v izidov kot druge. Boljše kot je ujemanje med strategijo, na primer ekonomskih teorijah odločanja, potrebno nadomestiti s pojmom določeno hevristiko, in strukturo naloge, bolj ekološko racionalni racionalnega vedenja, ki je kompatibilno z računskimi smo [35, 36]. sposobnostmi in dostopnostjo do informacij, kot jo ima človek v lastnem okolju v resnici. Racionalnost po njegovem ne pomeni iskanje najboljše možne, temveč zgolj dovolj dobre rešitve, kar 7 RACIONALNOST PREPRIČANJ je poimenoval satisficing. Uporabil je prispodobo škarij, kjer eno rezilo ponazarja računske zmožnosti akterja, drugo pa strukturo Poleg racionalnosti dejanj lahko govorimo tudi o racionalnosti okolja; zagovarjal je, da je pri preučevanju človeške prepričanj. Prepričanje je eden od temeljnih pojmov v racionalnosti pomembno upoštevanje in razumevanje obeh epistemologiji in je del klasične tripartitne definicije znanja, ki »rezil« [30, 31, 32, 33]. le-tega opredeli kot upravičeno resnično prepričanje. Eno od Simon je v svojih delih podrobno razdelal tako omejitve osrednjih vprašanj epistemologije je, kako priti do resničnih človekovega kognitivnega sistema kot značilnosti okolja. Menil prepričanj. Vprašanje je neločljivo povezano s preučevanjem je, da ni dokazov, ki bi pričali v prid temu, da človeško odločanje racionalnosti. Kakšen je odnos med racionalnostjo, poteka na način, kot to predpostavljajo normativne teorije, in da upravičenostjo in resničnostjo prepričanj ter znanjem? So ljudje v kompleksnih odločitvenih situacijah uporabljamo racionalna prepričanja tista, ki so upravičena, ali gre za ločena poenostavitve. Ena od njih je, da ne iščemo najboljše možne, pojma? Kako ljudje oblikujemo svoja prepričanja in kako bi jih optimalne rešitve, temveč si postavimo kriterij in izide nad njim morali [37, 38]? obravnavamo kot zadovoljive, pod njim pa kot nezadovoljive. Tradicionalni pogled je, da je vprašanje, kako bi ljudje morali Seveda se ob tem poraja vprašanje, na kakšen način si oblikovati prepričanja, v domeni epistemologije, vprašanje, kako postavljamo kriterij. Poleg tega pogosto nimamo popolnih dejansko jih, pa v domeni psihologije, in da naj bi disciplini informacij o tem, do kakšnih izidov bodo privedle različne delovali ločeno ena od druge. Do neke mere drži, da so alternative. Simon je zagovarjal, da v samem procesu odločanja normativna vprašanja epistemologije ločena od deskriptivnih postopoma pridobivamo informacije o tem in posodabljamo naše vprašanj psihologije - če bi določena psihološka spoznanja na poznavanje odnosa med alternativami in izidi. Vrednotenje primer pričala o tem, da je proces oblikovanja prepričanj alternativ po njegovem mnenju poteka postopoma, zaporedno, in pretežno nezaveden in da ljudje večinoma stremimo k tem, da odločevalec lahko preprosto izbere prvo zadovoljivo. Kriterij za sprejmemo prepričanja, ki spadajo v že obstoječo mrežo to, kaj je zadovoljiva rešitev, lahko po potrebi prilagajamo – če prepričanj, to samo po sebi ne daje dodatne teže koherentistični je previsok, ga znižamo in obratno, s čimer zagotovimo, da bomo teoriji upravičenja v epistemologiji. Vprašanji sta se začeli v vsakem primeru prišli vsaj do ene rešitve [30]. povezovati v 60. letih prejšnjega stoletja, ko je Willard V. O. Poleg zmožnosti organizma je za razumevanje racionalnosti Quine predstavil program naturalistične epistemologije, ki potrebno upoštevati tudi strukturo okolja. Simon je menil, da se poudarja, da so pri preučevanju prepričanj in znanja potrebne moramo osredotočiti na lastnosti okolja, ki so za odločevalca tudi metode, izsledki in teorije empiričnih znanosti [38]. pomembne in ki predstavljajo njegov življenjski prostor. Ne gre V literaturi se pogosto pojavlja izraz epistemska racionalnost. torej preprosto za preučevanje fizičnih lastnosti sveta, ki nas Pritchard jo opredeli kot obliko racionalnosti, katere cilj je obdaja; to, kaj smatramo kot okolje, je odvisno od zaznavnih pridobivanje resničnih prepričanj [37]. Po njegovem lahko sposobnosti, želja, potreb in ciljev organizma. Po Simonovem človek, ki stremi k epistemski racionalnosti, privzame različne mnenju odločevalci nimajo le enega, temveč več različnih strategije. Ena od njih je maksimizacija števila resničnih mehanizmov odločanja, ki so hierarhično urejeni, in vprašanje, prepričanj, druga pa minimizacija števila napačnih prepričanj, ki si ga moramo zastaviti, je, katere procese odločanja bomo v vendar pri obeh naletimo na težave: najboljši način za posameznih situacijah še lahko označili za prilagoditvene [31]. maksimizacijo števila resničnih prepričanj je, da verjamemo kar Vprašanje, kaj pomeni racionalno obnašanje, je torej koli, s čimer neizogibno pridobivamo tudi napačna prepričanja, drugačno, če ga zastavimo z upoštevanjem omejitev odločevalca najboljši način za minimizacijo števila napačnih prepričanj pa, in njegovega okolja ali pa iz perspektive normativnih teorij da ne verjamemo skoraj ničesar. Zdi se, da bi bilo najbolj racionalnosti. Ob upoštevanju vseh omejitev človeka, zlasti smiselno privzeti vmesen, uravnotežen pristop med verjetjem glede računskih in napovednih sposobnosti, je dejanska, vsemu in radikalnim skepticizmom [37]. Cilj epistemske človeška racionalnost lahko v najboljšem primeru le racionalnosti pa ni postavljen v prihodnost, temveč v sedanjost – poenostavljen približek t. i. globalne racionalnosti, na kateri večino epistemologov zanima, kakšno je stanje naših resničnih slonijo npr. modeli teorije iger [30]. prepričanj v tem trenutku, ne pa na primer čez eno leto. Za primer Koncept omejene racionalnosti se je od izvirnih Simonovih lahko vzamemo osebo, ki je brez ustreznih dokazov prepričana, del do danes razvijal in nadgrajeval ter še vedno močno vpliva da je dobra v matematiki. To prepričanje vodi v obiskovanje na preučevanje odločanja in racionalnosti [34]. Na njem temelji dodatnih ur matematike in zvišuje motivacijo ter količno učenja, tudi delo psihologa Gerda Gigerenzerja in sodelavcev, ki so kar na dolgi rok dejansko pripomore k večjemu številu resničnih osnovali raziskovalni program hitrih in varčnih hevristik ter so prepričanj o matematiki. Kljub temu bi večina epistemologov 30 zavrnila idejo, da je posedovanje prvega prepričanja epistemsko racionalnosti in epistemologije. Racionalnost prepričanj ali racionalno [38]. teoretsko racionalnost se pogosto prikazuje kot protipol praktični, Do zdaj omenjeni pogled prepričanja pojmuje kategorično, instrumentalni racionalnosti in menim, da je za razumevanje pri čemer imamo le tri možnosti: lahko smo prepričani, da p, celotne slike pomembno poznavanje obeh pogledov ali »vrst« prepričani, da ne-p, ali pa se prepričanja vzdržimo. Nekatera racionalnosti. Osredotočila sem se predvsem na odnos med področja epistemologije, na primer bayesovska epistemologija, racionalnostjo in različnimi temeljnimi pojmi epistemologije, pa prepričanja obravnavajo kot stopenjska – prepričanje torej ni zlasti upravičenjem, ter na povezavo med racionalnostjo in več propozicionalno stanje v smislu »vse ali nič«, temveč smo epistemskimi normami. Seveda pa na področju racionalnosti lahko v neko propozicijo prepričani bolj ali manj. V tem primeru prepričanj obstajajo še številna druga vprašanja in pogledi, opis se odpirajo številna nova vprašanja, na primer kakšen je odnos katerih presega namen prispevka. med dokazi za določeno propozicijo in našo stopnjo prepričanja vanjo ter kakšno stopnjo prepričanja potrebujemo, da lahko trdimo, da je posedovanje nekega prepričanja epistemsko 8 ZAKLJUČEK racionalno [39, 40, 41]. S tem povezana so tudi vprašanja o tem, Racionalnost je kompleksen pojem, ki zajema široko paleto kako prepričanja posodabljamo ali spreminjamo, ko človeškega spoznavanja in delovanja. Opredelitve racionalnosti, pridobivamo nove informacije. Obstajajo različni modeli, ki kriteriji zanjo in metode, s katerimi jo preučujemo, so tako opisujejo te procese, na primer AGM model revizije prepričanj številne in raznolike, da kategoričnega odgovora na vprašanje, [42] in teorija rangiranja [43, 44]. ali smo ljudje racionalni, ni pričakovati. Hkrati so praktično vsa Nadaljnja vprašanja, povezana z epistemsko racionalnostjo, področja našega življenja prepredena vsaj z implicitnimi se dotikajo epistemskih norm in odgovornosti. Pravila, ki nam prepostavkami o lastni (i)racionalnosti in tako je preučevanje le-narekujejo, kako oblikovati prepričanja, se imenujejo epistemske te pomembno ne le iz teoretskega, ampak tudi iz aplikativnega norme. Poraja se vprašanje, ali lahko agenta, ki prepričanja vidika. Preučevanje racionalnosti kot optimalnega doseganja oblikuje v skladu z napačnimi epistemskimi normami, še vedno ciljev lahko služi kot podlaga za oblikovanje spodbud in strategij, smatramo za epistemsko racionalnega. Šibkejši, deontični pogled ki bi tako posameznikom v vsakdanjem življenju kot na epistemsko racionalnost pravi, da ja – agentova prepričanja so strokovnjakom z različnih področij, kot so zdravstvo, epistemsko racionalna, če so v skladu z epistemskimi normami, gospodarstvo in pravo, pomagale pri učinkovitem sprejemanju ki jim agent sledi. V hipotetični situaciji, kjer bi bil agent dobrih odločitev. Tu pa pridemo do naslednjega vprašanja, ki se sistematično zaveden glede epistemskih norm, ni odgovoren za odpre, ko presežemo instrumentalno pojmovanje racionalnosti – morebitna napačna prepričanja; nasprotno pa v situaciji, kjer je kaj so »dobre« odločitve ali »racionalni« cilji? In nenazadnje, bil seznanjen s pravimi epistemskimi normami, pa vseeno sledi zakaj bi si pravzaprav želeli biti racionalni – ker menimo, da je napačnim, odgovornosti za napačna prepričanja ni razrešen. tako prav, ker racionalno delovanje izboljšuje naše možnosti za Močnejši, ne-deontični pogled pa kot kriterij za epistemsko preživetje in uspeh, ker vodi v srečo in blagostanje? Tudi pri racionalnosti postavlja, da agent sledi pravim epistemskim racionalnosti prepričanj se odpirajo podobna vprašanja; eno od normam, torej tistim, ki dejansko vodijo do resnice. Težava ne-njih je, ali je doseganje resnice vedno primarni epistemski cilj. deontičnega pogleda je v tem, da agent nikoli ni odgovoren za Pojmovanje racionalnosti je pomembno tudi pri razmislekih napačna prepričanja – če sledi napačnim epistemskim normam, o različnih vidikih zaupanja v znanost. Na kakšen način je sicer ni epistemsko racionalen, vendar tudi ni odgovoren za svoje znanje, ki ga pridobivamo z znanstveno metodo, drugačno od zmote [37]. znanja, ki ga pridobivajo laiki v vsakdanjem življenju? Koliko Predmet razprave je tudi vprašanje o odnosu med epistemsko prostora za napake in kolikšno stopnjo negotovosti je smiselno racionalnostjo in upravičenjem. Nekateri izraza »epistemsko dovoliti, ko preverjamo hipoteze? Kakšni dokazi so dovolj dobri, racionalna prepričanja« in »epistemsko upravičena prepričanja« da bomo neko trditev sprejeli ali ovrgli? Odgovori na ta in uporabljajo kot sinonima, drugi ju ločujejo. V drugem primeru ni podobna vprašanja so deloma odvisni od tega, kakšen pogled na jasno, kakšen je odnos med epistemsko racionalnim racionalnost privzamemo. prepričanjem in znanjem. Ena od možnih pozicij je, da tudi če Menim, da sta pri preučevanju racionalnosti pomembni tako sprejmemo upravičenje vsaj kot nujen, če ne že zadosten pogoj filozofska analiza kot metode empiričnih znanosti, ki nam dajejo za znanje, za epistemsko racionalnost to ne velja. Epistemsko vpogled v procese in mehanizme v ozadju človeškega racionalna prepričanja torej z znanjem niso povezana na enak oblikovanja prepričanj, sklepanja, presojanja in odločanja. način kot upravičena prepričanja. Če prekinemo povezavo med Integracija spoznanj različnih disciplin lahko pripomore k znanjem in epistemsko racionalnostjo, nam to omogoča, da zmanjševanju vrzeli med normativnimi in deskriptivnimi slednjo preučujemo tudi izven okvirja epistemologije in jo teorijami ter pripomore k oblikovanju karseda celostne slike povežemo z drugimi vidiki racionalnosti, na primer človeške racionalnosti. racionalnostjo odločitev in dejanj. Foley predlaga, da je odločitev (načrt, strategija) za osebo racionalna, če lahko oseba epistemsko racionalno verjame, da bo odločitev v zadovoljivi meri vodila v izpolnitev njenih ciljev [38]. Namen tega dela prispevka je bil nakazati le nekatera izmed številnih vprašanj, ki se odpirajo na presečišču preučevanja 31 VIRI [35] Gerd Gigerenzer and Peter M. Todd. 1999. Simple heuristics that make us smart. Oxford University Press, New York [1] Keith E. Stanovich. 2010. Decision making and rationality in the modern [36] Gerd Gigerenzer, Ralph Hertwig and Thorsten Pachur (ur.). 2011. world. Oxford University Press, New York Heuristics: the foundations of adaptive behavior. Oxford University Press, [2] Phillip E. Tetlock and Barb Mellers, 2002. The great rationality debate. New York Psychol Sci 13, 1 (Jan, 2002), 94–99. DOI: 10.1111/1467-9280.00418 [37] Duncan Pritchard. 2018. What Is This Thing Called Knowledge? (4th [3] Edward Stein. 1996. Without good reason: the rationality debate in edition.) Routledge: New York philosophy and cognitive science. Clarendon Press, Oxford [38] Sven Bernecker and Duncan Pritchard (Ed). 2011. The Routledge [4] Ronald de Sousa. 2007. Why Think?: Evolution and the Rational Mind. Companion to epistemology. Routledge, New York Oxford University Press, New York [39] Richard Foley, 1992. The epistemology of belief and the epistemology of [5] Jon Elster. 1983. Sour Grapes: studies in the subversion of rationality. degrees of belief. Am Philos Q 29, 2 (Apr, 1992), 111–124 Cambridge University Press, New York [40] Stephen Hartmann and Jan Sprenger, 2011. Bayesian epistemology. In The [6] Wolfgang Köhler. 1925. The Mentality of Apes. Percy Lund, Humphries routledge companion to epistemology. Routledge, New York, 636–648 & Co, London [41] Elizabeth G. Jackson, 2020. Relationship between belief and credence. [7] John R. Searle. 2001. Rationality in Action. MIT Press, London Philos Compass 15, 6 (Mar, 2020), 1–13. DOI: 10.1111/phc3.12668 [8] Robert Nozick. 1981. The Nature of Rationality. Princeton University [42] Carlos E. Alchourrón, Peter Gärdenfors and David C. Makinson (1985). Press, New Yersey On the logic of theory change: Partial meet contraction and revision [9] Shaun H. Heap. 1989. Rationality in economics. Blackwell Publishing, functions. J Symb Log 50, 2 (June, 1985), 510–530, DOI: London 10.2307/2274239 [10] Peter C. Wason. 1966. Reasoning. In New horizons in psychology, Pelican, [43] Wolfgang Spohn, 1988. Ordinal conditional functions: A dynamic theory Harmondsworth of epistemic states. In Causation in decision, belief change, and statistics [11] Peter C. Wason, 1968. Reasoning about a rule. Q J Exp Psychol. 20, 3 II. Dordrecht: Kluwer, 105–134 (Apr, 1968), 273–281. DOI: 10.1080/14640746808400161 [44] Franz Huber, 2019. Ranking theory. In The open handbook of formal [12] Richard A. Griggs and James R. Cox, 1982. The elusive thematic-epistemology. PhilPapers Foundation, 397–437 materials effect in Wason’s selection task. Br J Psychol (July, 1982) 73, 407–420. DOI:10.1111/j.2044-8295.1982.tb01823.x [13] Marco Ragni, Ilir Kola and Phillip N. Johnson-Laird. The Wason selection task: A meta-analysis. In Proc 32nd Annu Conf Cogn Sci Soc (July, 2017). Curran Assciates Inc., New York, 980–985 [14] Phillip N. Johnson-Laird and Peter C. Wason, 1970. A theoretical analysis of insight into a reasoning task. Cogn Psychol 1, 2 (Feb, 1970), 134–148. DOI:10.1016/0010-0285(70)90009-5 [15] Valerie A. Thompson, Jonathan St. B. T. Evans and Jamie I. D. Campbell, 2013. Matching bias on the selection task: It’s fast and feels good. Think Reason (Oct, 2013), 19, 3–4, 431–452. DOI:10.1080/13546783.2013.820220 [16] Paul Pollard and Jonathan St. B. T. Evans, 1987. Content and context effects in reasoning. AJP 100, 1 (Spring, 1987), 41–60. DOI: 10.2307/1422641 [17] Leda Cosmides and John Tooby, 1992. Cognitive adaptations for social exchange. In The adapted mind: Evolutionary psychology and the generation of culture. Oxford University Press, New York, 163–228 [18] Paul S. Davies, James H. Fetzer and Thomas R. Foster, 1995. Logical reasoning and domain specificity - A critique of the social exchange theory of reasoning. Biol Philos 10 (Jan, 1995), 1–37. DOI:10.1007/BF00851985 [19] Mike Oaksford and Nick Chater, 1994. A rational analysis of the selection task as optimal data selection. Psychol Rev 101, 4 (Oct, 1994), 608–631. DOI:10.1037/0033-295X.101.4.608 [20] Mike Oaksford and Nick Chater, 2003. Optimal data selection: Revision, review, and reevaluation. Psychon Bull Rev 10, 2 (May, 2003), 289–318. DOI: 10.3758/BF03196492 [21] Baruch Fischhoff, 2010. Judgment and decision making. Wiley Interdiscip Rev Cogn Sci 1, 5 (Sept/Oct, 2010), 724–735. DOI: 10.1002/wcs.65 [22] John von Neumann and Oskar Morgenstern. 1944. Theory of games and economic behavior. Princeton University Press, New Yersey [23] Daniel Kahneman and Amos Tversky, 1979. Prospect theory: An analysis of decision under risk. Econometrica 47 (Mar, 1979), 263–291. DOI: 10.2307/1914185 [24] Daniel Kahneman, Amos Tversky and Paul Slovic. 1982. Judgment Under Uncertainty: Heuristics and Biases. Cambridge University Press, Cambridge [25] Daniel Kahneman and Amos Tversky. 2000. Choices, values and frames. Cambridge University Press, Cambridge [26] Daniel Kahneman. 2011. Thinking, fast and slow. Farrar, Straus and Giroux, New York [27] Aron K. Barbey and Steven A. Sloman, 2007. Base-rate respect: From ecological rationality to dual processes. Behav Brain Sci 30, 3 (Jun, 2007), 255–297. DOI: 10.1017/S0140525X07001653 [28] Gerd Gigerenzer, 1996. The psychology of good judgment: Frequency formats and simple algorithms. Med Decis Mak 16, 3 (Jul–Sept, 1996), 273–280. DOI: 10.1177/0272989X9601600312 [29] Gerd Gigerenzer, 1991. How to make cognitive illusions disappear: Beyond “heuristics and biases.” Eur Rev Soc Psychol. 2 , 1 (1991), 83–115. DOI: 10.1080/14792779143000033 [30] Herbert A. Simon, 1955. A behavioral model of rational choice. Q J Econ 69, 1 (Feb, 1955), 99–118. DOI: 10.2307/1884852 [31] Hebert A. Simon, 1956. Rational choice and the structure of the environment. Psychol Rev 63, 2 (Mar, 1956) 129–138. DOI: 10.1037/h0042769 [32] Herbert A. Simon, 1990. Invariants of human behavior. Annu Rev Psychol 41 (Feb, 1990), 1–19. DOI:10.1146/annurev.ps.41.020190.000245 [33] Herbert A. Simon, 1992. What is an explanation of behavior? Psychol Sci 3, 3 (May, 1992), 150–161. DOI: 10.1111/j.1467-9280.1992.tb00017.x [34] Riccardo Viale (ur.). 2021. Routledge handbook of bounded rationality. Routledge, New York 32 Joint history of play provides means for coordination Liubov Voronina† Christophe Heintz MEi:CogSci Department of Cognitive Science University of Ljubljana Central European University Ljubljana, Slovenia Vienna, Austria voronina.liuba@gmail.com HeintzC@ceu.edu ABSTRACT In pure coordination games (Schelling games) participants are asked to choose the same option from the set of equally attractive ones. Surprisingly, people tend to converge on one In this study we investigate how joint history shapes strategic particular option at a rate significantly higher than chance. For decisions for solving coordination problems. We show that example, choosing between «heads» or «tails» reveals coordinating partners use the history of their past interactions to consistent preference for «heads», much higher than select their strategies. More precisely, people accurately predict mathematically implied equiprobability. Such recognisable that a winning strategy used in the past is mutually salient and prominence of one alternative over another, that results in a can be successfully used again in similar situations. Thus, joint stable solution, is called a focal point or salience. history helps players form accurate mutual expectations about each others’ choices and increase the rate of successful Pure coordination games therefore pose a question how to coordination. identify a unique solution to avoid coordination failure [4]. Although the exact reasoning behind the coordination process is We demonstrate that precedence is strongly relied upon and open for debate [5, 6, 7, 8], Schelling’s suggestion is to look for provides insights into the psychological bases of the social such selection rule among many, which can single out a processes through which conventions emerge. By investigating successful coordination strategy. This rule should be mutually the path dependence of the individual behaviour in the context recognised by the interacting parties to be able to provide of coordination, we experimentally confirm that conventions reliable means for coordination [1]. A focal point, emerged by emerge because people systematically rely on their past applying such selection rule, is called Schelling salience. interactions in order to coordinate successfully. Building on the logic of coordination games, David Lewis KEYWORDS convincingly argued for the emergence of (linguistic) conventions [9]. According to his account, observed coordination games, path dependence, Schelling salience behavioural regularities that are commonly known among the population, create accurate mutual expectations that facilitate 1. INTRODUCTION coordination by providing unambiguous solution to social coordination problems, resulting in stable equilibria. Coordination is the process of tacit convergence on a mutual We hypothesise that these behavioural regularities become strategy in the context of interdependent decisions. salient by virtue of repeating precedence, which is used as Coordinating partners can choose to do exactly same thing Schelling salience, once the agents are confronted with the (drive on the right side of the road), exactly the opposite thing coordination problem. (wait while another person is calling back after the line is cut) or complement each other’s actions to produce a common The goal of the study is to show how the joint history of outcome (division of the household chores). In many everyday interactions in coordination problems shapes the choice of cases coordination is achieved by following an existing coordination strategies. At the cognitive level, this means that convention, by making an explicit verbal agreement or by joint history is used by people as relevant information for performing actions in sequence, when the person initiating the choosing their strategy for coordination. interaction has the opportunity to make the first choice and The following hypotheses were tested: express their preference. However, even in the absence of communication, successful coordination can be accomplished H1: Joint history facilitates accurate mutual expectations. with the probability much higher than chance. Thomas Players choose a coordination strategy in view of what they Schelling first draw attention to this apparent paradox of expect their partner to do. These expectations are informed by coordination with his informal experiments [1], that were later the knowledge they have of their joint history, which makes successfully replicated in the controlled settings [2, 3]. their prediction more accurate. H2: Joint history determines coordination strategies. When the situation does not provide any unambiguous clues for Permission to make digital or hard copies of part or all of this work for personal or coordination, players choose a specific strategy that resulted in classroom use is granted without fee provided that copies are not made or successful coordination in the past to resolve the ambiguity and distributed for profit or commercial advantage and that copies bear this notice and avoid coordination failure. the full citation on the first page. Copyrights for third-party components of this work must be honored. For all other uses, contact the owner/author(s). Information Society 2022, 10–14 October 2022, Ljubljana, Slovenia © 2022 Copyright held by the owner/author(s). 33 Information Society 2022, 10–14 October 2022, Ljubljana, Slovenia L.Voronina et al. 2. METHOD 2.2. Experimental conditions This research is based on the empirical methodology of We used a between-subject design to examine the research experimental game theory. The economic game chosen for the hypotheses. Participants were presented with the experimental experiment is pure coordination game [1, 2, 3]. scenarios under the following three conditions: In the experiment, participants were presented with various (1) no joint history (baseline condition) layouts of coloured tokens and asked to coordinate on the token Baseline data reveals the rate of coordination in the absence of of the same colour. Both sets of tokens were visible to both joint history of play. The baseline condition also helps to partners and the choice was simultaneous. The result of every empirically differentiate various types of rounds that are used to interaction and individual players' choices were logged online construct scenarios in the experimental conditions. Rounds that in real time. The analysis was carried out for the particular type reveal preference for one particular colour will become history of rounds (at individual or dyadic level) with the condition as rounds. Rounds, where the colour choices are equally an independent variable. distributed, will become test rounds. Coordination index [2, 3] 2.1. Participants is calculated to show the hypothetical coordination rate of the unpaired participants based on their individual responses. One hundred and thirty-three participants took part in the Experimental setting: Participants play single rounds in pairs “Mobile Coordination Games” experiment, which was with their player IDs hidden. They connect with the new partner conducted online in two parts. Game sessions for the baseline after each round and are aware of this. The participants pool is condition were organised during June and July, 2021 with a set to 2 players to allow random pairing. total of 51 participants (mean age = 26.2 years; 16 females and 35 males). Game sessions for the experimental conditions took Experimental stimuli: 30 individual rounds of various difficulty. place in January, 2022 with a total of 82 participants (mean age Each round could be played for up to 2 times by any player (but = 24.1 years; 24 females and 58 males). Participants were not in sequence). recruited online via the Sona Research Participation System of (2) random joint history (experimental condition) Central European University. There were no restrictions on Participants have the opportunity to build a joint history of play, participation for the adult participants, who needed basic consisting of randomly assigned rounds. This history of mutual English skills for understanding the instructions and a mobile interactions can provide them with the clues for successful device for accessing the Coordy research application. All the coordination in the test round. Its coordination rate will be participants received compensation based on their performance compared to the corresponding baseline rate and the level (average amount = 4,9 euros) in the form of an online coordination index. voucher of their preference, either Amazon or PayPal. Experimental setting: Participants play games, consisting of 2.2. Materials four random rounds and a test round, in dyads with their player IDs shown. They change their game partner after each game. To enable empirical investigation of the real-time coordination The participants pool is set to 8 players to allow fixed pairing in between the pairs of participants, a proprietary mobile research order to avoid repetitions. application named Coordy has been developed for both Android and iOS based mobile devices. Coordy was officially released Experimental stimuli: 36 games of 5 rounds from the baseline and became available for download on Google PlayMarket and condition (6 unique histories of four rounds combined with AppStore. each of the 6 test rounds). Each game was played just once In the experiment, we used two different kinds of experimental during the game session. scenarios: (3) specified joint history (experimental condition) • 30 single rounds of various difficulty in the baseline Joint history, provided by the designed scenario, increases the condition: probability that a certain strategy is used during this history A. easy rounds with the symmetrical clues for coordination; and, subsequently, in the test round. Individual player’s B. hard rounds with the clashing clues; strategy, operationalised as a choice of the specific colour, will C. equiprobable rounds with no coordination clues; be compared between different histories that end up with the • 46 games of five rounds (experimental conditions). same test round. Experimental setting: Participants play games, consisting of four predefined rounds and a test round, in dyads with their player IDs shown. They change their game partner after each game. The participants pool is set to 8 players to allow fixed pairing in order to avoid repetitions. Experimental stimuli: 10 games of 5 rounds (5 unique histories of four rounds combined with the corresponding test rounds). Each game was played twice during the game session. All the scenarios (both single round and games of five rounds) appeared during the game sessions in the randomised order to avoid order effects. The order of rounds within a particular game was fixed. Both experimental conditions were tested Figure 1: Examples of A, B and C rounds. together during the same game sessions. 34 Joint history of play provides means for coordination Information Society 2022, 10–14 October 2022, Ljubljana, Slovenia 3. RESULTS Participants were more likely to successfully coordinate in the test rounds after the joint history of play than without it. Before reporting the results of the study, we would like to clarify the issue of the players’ expertise and its potential influence on the outcome of coordination. In both conditions, all the participants would start playing without any prior experience (match number 0). We analysed the outcome of the coordination in the last round (success or failure) for match numbers below and above 7 (half of the experimental game sessions) and found no evidence for the improvement of the coordination success at the dyadic level. Coordination in the last round was successful in about half of the games irrespective of the participants’ level of experience with the task. 3.1. Baseline results A chi-square test of goodness-of-fit was performed to determine Figure 3: Change in the coordination rate and coordination whether each of the three colours were equally chosen by the index across conditions (group-level). Levels of significance: participants in the particular round. **: p < .01, ***: p < .001. A two-proportion z-test was conducted to calculate the difference between the CR and coordination index (CI) in the last round of the games with random history. For the group of test B rounds CR was found to be significantly higher than CI after the joint history of play z (N = 1029 ) = 4.26, p < .001; r = .13. Therefore the actual coordination rate exceeds the rate of the expected coordination, when the choices are made by the randomly paired participants. 3.3. Specified history results Figure 2: Baseline coordination rates for the A, B rounds. A chi-square test of independence was performed to examine the relation between the individual player’s choice in the same A preference for a specific colour was found in the majority of test round and the specific history preceding that round. A rounds (except for A15) and some B rounds (B3, B11, B12). The corresponding dots on the graph are coloured with the colour that was chosen the most (over ⅔ of the individual choices) in the particular round. Also the coordination rates (CRs) for those rounds were very high (mean CR = 0.72). These rounds were used to constitute history rounds. A preference for the specific colour was not found in the four B rounds (B1, B8, B17, B18) and one A round (A15). While the choices for the three colours were not equally distributed in the rounds B2, B4, B7 and B9, the proportion of any particular colour did not exceed 60%. Their corresponding dots on the graph are therefore coloured in black. Also their CRs were significantly lower (mean CR = 0.43) than in the previous Figure 4: Individual player’s choice of colour for group of rounds with the focal points. These rounds were used coordination in the last round after the specified history. as test rounds in the random history condition. For the majority of C rounds (C1, C3, C4, C6) a colour For some pairs of histories, the relation between these variables preference was not established. Also the CRs for C rounds were was significant: not significantly higher than chance (mean CR = 0.57). Hence • for the test round C4 after the histories Red and Blue they are not depicted on the graph. These rounds were used as X² (1, N = 74) = 7.29, p < .01; r = .3; test round in the specified history condition. • for the test round C2 after the histories Green and Blue X² (1, N = 74) = 15.69, p < .0001; r = .45. 3.2. Random history results In the same test round, participants were more likely to choose A chi-square test of independence was performed to examine the modal colour of the history rounds (e.g. in the test round C4 the relation between the coordination rate in the test rounds (at participants were more likely to choose red colour after the the group level) and the history of previous interactions. The history Red and blue colour after the history Blue). relation between these variables was significant, X² (1, N = 592) For the histories Red and Green the relation between these = 8.39, p < .01; r = .12. variables was statistically insignificant. 35 Information Society 2022, 10–14 October 2022, Ljubljana, Slovenia L.Voronina et al. 4. DISCUSSION Overall, in our experiment we managed to observe how participants make use of the precedence by applying the In this study we aimed to investigate how previous interactions following rule for coordination: «choose the colour that brought can influence the outcome of coordination for the pair of us successful coordination before». Though studying path players. First, we let the participants play single rounds dependence in the lab setting poses certain challenges [10], anonymously. Even though the participants played several some researchers found the way to address them using the rounds, they could not constitute joint history of interaction economic games [11]. In the future, it could be fruitful to because they were — knowingly — paired with a new random empirically investigate the robustness of the coordination rules participants for each round. This set-up helped us identify and the amount of common knowledge required for their rounds with the «natural» focal points, i.e. colours that appealed emergence [12, 13, 14, 15]. to the participants as obvious to coordinate upon due to the specific layout of the scenario, irrespective of other factors. ACKNOWLEDGMENTS We noticed that in the absence of communication and any explicit coordination rules, participants did manage to The authors acknowledge financial support by the European coordinate more than rational choice theory would predict. This Research Council, under the European Union‘s Seventh is in line with previous results showing that people are able to Framework Programme (FP7/2007-2013)/ERC grant agreement rely on Schelling salience in order to coordinate successfully. no. 609819 (Somics project). In our experiment participants converged on a tacit rule for REFERENCES coordination, which was «choose colour with the most tokens 1. Thomas Schelling. 1960. The strategy of conflict (1st. ed.). Harvard present on both players’ layouts». Those rounds, where this rule University Press, Cambridge, MA. could not be unmistakably applied, demonstrated lower 2. Judith Mehta, Chris Starmer, and Robert Sugden, 1994. The Nature of coordination rates and were chosen to be the test rounds for the Salience: An Experimental Investigation of Pure Coordination Games. subsequent experimental conditions. We wanted to explore the The American Economic Review 84, 3 (Jun, 1994), 658–73. JSTOR: http://www.jstor.org/stable/2118074. possibility that the history of interactions itself would provide 3. Nicholas Bardsley, Judith Mehta, Chris Starmer and Robert Sugden, Schelling salience and thus determine the choice of colour to 2010. Explaining Focal Points: Cognitive Hierarchy Theory versus Team increase the coordination rate. Reasoning, The Economic Journal 120, 543 (Mar, 2010), 40–79. DOI:https://https://doi.org/10.1111/j.1468-0297.2009.02304.x. We then created games with five rounds, which were played by 4. Giovanna Devetag and Andreas Ortmann, 2007. When and why? A the participants in dyads with their IDs shown and mutually critical survey on coordination failure in the laboratory. Experimental known, thus letting them build the history of mutual economics 10 (Aug, 2007), 331-344. DOI:https://doi.org/10.1007/ s10683-007-9178-9. interactions. In the games, which histories did not suggest a 5. Robert Sugden. The logic of team reasoning, 2003. Philosophical choice of any specific colour, we observed a significant explorations 6, 3 (Sep, 2003), 165-181. DOI:https://doi.org/ improvement coordination in the last round. Interestingly, the 10.1080/10002003098538748. coordination index for this round did not significantly changed 6. Colin F. Camerer, Teck-Hua Ho and Juin-Kuan Chong, 2004. A Cognitive Hierarchy Model of Games, The Quarterly Journal of between the conditions with and without joint history. It is only Economics 119, 3 (Aug, 2004), 861–898. DOI:https://doi.org/ the actual coordination rate that changed. In other words, had 10.1162/0033553041502225. the participants been paired randomly for the last round, no 7. Nicholas Bardsley and Aljaž Ule, 2017. Focal points revisited: Team improvement would have occurred. reasoning, the principle of insufficient reason and cognitive hierarchy theory. Journal of Economic Behavior & Organization 133 (Jan, 2017), This suggests that the increase in coordination rate is due to 74-86. DOI:https://doi.org/10.1016/j.jebo.2016.10.004. players tracking what they have played with their own partner 8. Francesco Guala, 2018. Coordination, Team Reasoning, and Solution and using this information to make their future choices. This Thinking. Revue d'économie politique 128, 3, 355-372. DOI:https:// doi.org/10.3917/redp.283.0355. findings confirm our hypothesis that joint history of play 9. David Lewis. 1969. Conventions: A Philosophical Study (1st. ed.). facilitates coordination. When the game partners are aware of Harvard University Press, Cambridge, MA. each other’s previous choices, they tend to choose the focal 10. Jean-Philippe Vergne and Rodolphe Durand, 2010. The missing link point for coordination more accurately. However, the effect size between the theory and empirics of path dependence: conceptual clarification, testability issue, and methodological implications, Journal of the observed differences remained small. One possible of Management Studies 47, 4 (Aug, 2010), 736-759. DOI:https://doi.org/ explanation is that randomness and variety of the history rounds 10.1111/j.1467-6486.2009.00913.x created clashing focal points to converge on. 11. Marc Knez and Colin Camerer, 2000. Increasing cooperation in prisoner's dilemmas by establishing a precedent of efficiency in In the games from the third condition, where a history of rounds coordination games, Organizational Behavior and Human Decision nudged the choice of a given colour, we observed that this same Processes 82, 2 (Jul, 200), 194-216. DOI:https://doi.org/10.1006/ colour tended to be chosen in the last round. More precisely, obhd.2000.2882 12. Margaret Gilbert, 1983. Agreements, conventions, and language. participants had to select one of two colours in the last round of Synthese 54, 3 (Mar, 1983), 375-407. JSTOR: http://www.jstor.org/ their joint history. They tended to select the same colour, on stable/20115846. which they coordinated during their joint history. They did so 13. Ken Binmore, 2008. Do conventions need to be common significantly more than the participants, who were given a joint knowledge? Topoi 27 (Jul, 2008), 17-27. DOI:https://doi.org/10.1007/ s11245-008-9033-4. history that nudged towards another colour. We documented 14. Giacomo Sillari, 2008. Common Knowledge and Convention. Topoi 27 that effect for two test rounds with a moderate effect size. We (Jul, 2008), 29-39. DOI:https://doi.org/10.1007/s11245-008-9030-7. did not observe a significant effect for the third test round. 15. Kyle A. Thomas, Peter DeScioli, Omar Sultan Haque, and Steven Pinker, 2014. The Psychology of Coordination and Common Knowledge, Our post-hoc hypothesis is that the salience of the red colour Journal of Personality and Social Psychology 107, 4, 657–676. overshadowed the salience shaped by the history of past DOI:https://doi:10.1037/a0037037. interactions. This mixed result calls for the replication study with the different set of stimuli. 36 Predicting Trust in Science in the Context of COVID-19 Information Society 2022, 10–14 October, Ljubljana, Slovenia Predicting Trust in Science in the Context of COVID-19 Pandemic: The Role of Sociodemographics and Social Media Use Žan Zelič Martin Berič Darja Kobal Grum Department of Psychology Department of Psychology Department of Psychology University of Ljubljana University of Ljubljana University of Ljubljana Ljubljana, Slovenia Ljubljana, Slovenia Ljubljana, Slovenia zan.zelic@gmail.com martinberic@yahoo.com darja.kobal@ff.uni-lj.si ABSTRACT will perform an assigned task with a certain level of competence and expertise, while moral norms are related to the anticipation Research in the context of COVID-19 pandemic has consistently that by doing so, they will also act in a way that puts the interest shown that scientific distrust adversely affects health-related of the community before their personal advances. Similarly, behavior. Therefore, the aim of our study was to identify the risk Wintterlin et al. [4] argue that trust in science is rooted in factors for the development of scientific distrust, with emphasis expectations that scientists’ claims are epistemically sound and on the role of sociodemographic variables and social media use. that science has a prosocial stance. Overall, perhaps the most A convenience sample of 490 Slovene speaking individuals was comprehensive definition of trust in science has been provided used to perform hierarchical linear regression analysis. In line by Nadelson et al. [5], describing it as a multifaceted construct, with our hypotheses, the results showed that trust in science was which includes affective components, credibility and negatively correlated with age, religiosity and use of social media trustworthiness perceptions, knowledge and epistemic beliefs. as an information source about COVID-19, while it was Since scientists were the main source of information on positively correlated with male gender and total years of formal COVID-19 and its adverse consequences, and also the ones that education. When only sociodemographic variables were entered helped governments develop preventive measures and vaccines, into the prediction model, each of them explained a significant the findings that low trust in science negatively impacts health-proportion of the variance in trust in science. However, after the related behavior [2] should not come as a complete surprise. inclusion of social media use, religiosity was no longer a However, not much has been researched about the predictors of significant predictor. In contrast to our expectations, the results trust in science in the context of the pandemic. As we believe this also showed no significant interaction between education and kind of knowledge is crucial to implement communication social media use when predicting trust in science. Our findings changes, which could accurately address those who are are further discussed and additional implications are provided. particularly prone to developing scientific mistrust, we conducted a study focusing on the sociodemographic predictors KEYWORDS of trust in science as well as its connection to social media use. COVID-19, trust in science, social media use, education, religiosity 1.1 Predictors of Trust in Science Previous research on the relationship between trust in science and 1 INTRODUCTION age has shown somewhat mixed results. For example, some researchers reported on non-significant correlations [6], while When the new coronavirus (Sars-CoV-2) started to spread in others found that scientists were more likely to be trusted by 2020 it has quickly become evident that the world as we knew it those who are younger [7]. The latter result could in part be was about to change. Ever increasing number of infections led to explained by higher average levels of education among younger health system overloads, high mortality rates, mental health individuals, however age remained an important predictor even difficulties and great economic burden [1]. As adoption of social when education was accounted for [7]. distancing measures and newly developed vaccines was crucial Regarding gender, previous research has consistently shown for reducing the spread of the new coronavirus and its adverse that men generally have more positive attitudes towards science consequences, identification of factors influencing health-related than women [6][8]. However, when possible reasons for these behavior became of utmost importance. One of the variables that results were examined, other sociodemographic variables, such has been consistently found to predict preventive behavior as as education, religiosity and work status were found to explain well as vaccine acceptance is trust in science [2]. this gender gap [8]. According to Barber [3] public trust in science depends on the Throughout history, religion and science were often seen as perceptions of scientists’ compliance with technical and moral epistemologically conflicted [9], which may have resulted in norms. Technical norms consist of expectations that the scientists lower trust in science by those who are more religious. Indeed, previous research has shown that religiosity was associated with Permission to make digital or hard copies of part or all of this work for personal or classroom use is granted without fee provided that copies are not made or distributed negative attitudes towards science as well as lower science for profit or commercial advantage and that copies bear this notice and the full literacy [5][10]. citation on the first page. Copyrights for third-party components of this work must be honored. For all other uses, contact the owner/author(s). Another sociodemographic factor that has been consistently Information Society 2022, 10–14 October 2022, Ljubljana, Slovenia shown to predict trust in science is education [6]. One of the most © 2022 Copyright held by the owner/author(s). prevalent explanations for the described relationship was that 37 Information Society 2022, 10-14 October, Ljubljana, Slovenia Ž. Zelič et al. education indirectly influences positive attitudes towards science 0 and 100 to estimate the percentage of information about the by increasing scientific knowledge [11]. However, further new coronavirus they obtained through social media. research showed that education remained an important predictor Trust in Science was measured by the Trust in Science and of trust even when controlling for scientific knowledge [12]. Scientists Inventory [5], which contains 21 items (e.g., We can Although previous research has indicated that social media trust science to find answers that explain the natural world. ). use positively predicts trust in science [13], we believe that the Participants were asked to rate their agreement with the provided results might be different in the times of COVID-19 pandemic. statements on a 5-point Likert scale with anchors, 1 ( Strongly Since social platforms enabled rapid misinformation dispersion disagree) and 5 ( Strongly agree). Confirmatory factor analysis [14], extensive social media use could lower trust in science by (CFA) showed poor one-factor model fit, so we excluded item increasing conspiracy beliefs about scientists’ involvement in the 11, which was semantically very similar to items 9 and 10. pandemic. Indeed, our previous research [15] showed that the Additionally, we allowed for some residual covariances extent of using social media as an information source predicted according to modification indexes. The fit of the modified 20-COVID-19 conspiracy beliefs, which were also highly inversely item scale was acceptable: χ 2(166) = 484.642, p < .001, CFI correlated with trust in science. = .939, TLI = .930, RMSEA = .070, 90% CI: [.063, .078], SRMR = .042. The shorter version of the scale also showed excellent 1.2 The Present Research internal consistency ( α = .95). The aim of our study was to examine the importance of several sociodemographic variables and social media use in predicting 3 RESULTS trust in science. Based on the previous findings we hypothesized that trust in science will be higher among younger individuals Firstly, the factor structure of the translated Trust in Science and (H1), men (H2), those who are less religious (H3), more educated Scientist Inventory was assessed by confirmatory factor analysis (H4) and those who obtained less information about the (CFA), using R package lavaan [17]. Since the data were non-coronavirus from the social media (H5). Additionally, we normally distributed, we used the robust maximum likelihood hypothesized that education would have a moderating role in the method (MLM) of model estimation. After minor modifications relationship between social media use and trust in science (H6). were implemented to achieve an acceptable one-factor model fit, Since critical thinking has been found to develop through the total trust in science score was calculated as a mean value of education [16], we assumed that even extensive social media use all items. All further analyses were done in IBM SPSS version would not reflect in high levels of scientific distrust as long as 25.0 [18]. individuals would be capable to critically evaluate the quality of Secondly, descriptive statistics and intercorrelations were obtained information. Furthermore, we also aimed to investigate calculated for all measured variables. The results showed that the amount of the variance in trust in science that a combination trust in science was negatively correlated with age ( r = -.14, p of these variables could explain as well as their relative = .002), religiosity ( r = -.16, p < .001) and use of social media as importance when entered into a multivariate prediction model. an information source about COVID-19 ( r = -.35, p < .001), while it was positively correlated with male gender ( rpb = .21, p < .001) and total years of formal education ( r = .29, p < .001). 2 METHOD Thirdly, when we determined that all assumptions for multiple linear regression were met, hierarchical linear 2.1 Sample regression analysis was conducted. Trust in science was entered Data collection took place between March 29 and April 7, 2021, into the analysis as a criterion variable, while all other measured using an online survey. Convenience sample was used, variables were consecutively added as predictors (see Table 1). consisting mostly of students at the University of Ljubljana and In the first step age and gender together explained 6.5% of members of different COVID-19 related Facebook groups. variance in trust in science with younger individuals and men Responses of 490 participants (397 women, 92 men and one non-exhibiting more trust. Both predictors were significant, although binary), aged from 18 to 70 years ( M = 35.7, SD = 13.2), were the relative importance of gender was greater. In the second step analyzed. The majority (56.5%) of the participants had a college religiosity explained only 1.8% of additional variance in trust in degree, 41.8% reported on having a high school diploma and science, however the change in R2 was statistically significant. 1.6% completed only elementary school. Furthermore, 31.6% of Those who were less religious showed significantly higher levels them were students, 54.7% were employed, 9.0% were of trust even when age and gender were accounted for. unemployed and 4.7% were retired. Furthermore, both age and gender remained significant predictors of trust despite slight decrease of gender's b value. In 2.2 Measures the third step the years of formal education turned up to be the Demographic data was obtained through a series of questions on most important positive predictor of trust in science, additionally age, gender, years of education and employment status. explaining 6.7% of its variance. Inclusion of education slightly Religiosity was measured by the participants’ level of lowered the b values of age and religiosity, however all included agreement with the statement “I would define myself as a predictors remained statistically significant. In the fourth step the religious person.” on a 7-point Likert scale with anchors, 1 share of COVID-19 information obtained from social media was ( Strongly disagree) and 7 ( Strongly agree). added into the equation, explaining an additional 5.8% of Use of social media as an information source about COVID-variance in trust in science. The results showed that those who 19 was measured by moving an interactive slider between values relied more on social networks to obtain information were less likely to trust in science. Altogether, a combination of five 38 Predicting Trust in Science in the Context of COVID-19 Information Society 2022, 10–14 October, Ljubljana, Slovenia Table 1: The results of the hierarchical linear regression analysis Step 1 Step 2 Step 3 Step 4 Step 5 Variable B SE b B SE b B SE b B SE b B SE b Age -.01 .00 -.14 ** -.01 .00 -.14 ** -.01 .00 -.11 ** -.01 .00 -.11 ** -.01 .00 -.11 ** Male gender .46 .10 .21 *** .42 .10 .19 *** .41 .09 .19 *** .28 .09 .13 ** .28 .09 .13 ** Religiosity -.07 .02 -.14 ** -.05 .02 -.11 * -.03 .02 -.07 -.03 .02 -.07 Education .10 .02 .26 *** .08 .02 .21 *** .09 .03 .22 ** SM information -.01 .00 -.26 *** -.01 .01 -.21 Edu x SM info .00 .00 -.05 R2 .064 .083 .149 .207 .207 △ R2 .064 *** .018 ** .067 *** .058 *** .000 Note. * p < .05. ** p < .01. *** p < .001. predictors explained 20.7% of variance in trust in science. significant predictor of trust in science even when age, religiosity However, after the variable of social media use was included, b and years of education were controlled. One possible explanation values of gender, education and religiosity decreased, thus for this result may be that on average women have less specific designating religiosity as a non-significant predictor. Finally, science-related knowledge than men. Although in our research analysis in the fifth step showed that there was no significant male gender was not significantly related to years of total interaction between education and social media use when education ( rpb = .04, p = .440), education of men and women predicting trust in science. might differ in terms of its type and field of interest. For example, Global Gender Gap Report 2022 showed that only 33% of STEM graduates in Slovenia are female [23]. In line with the above, Fox 4 DISCUSSION AND CONCLUSIONS & Firebaugh [24] also found that years of education did not The aim of our research was to examine the predictors of trust in explain the gender gap in science confidence. Moreover, their science in the context of COVID-19 pandemic since such research pointed out that gender differences can in large part be knowledge could be used to implement communication changes attributed to lower perceived utility of science by women. that might motivate higher compliance with preventive measures Based on previous studies, which showed that religiosity and protect public health. predicts negative attitudes towards science [10], we also Regarding age, the results were in line with our assumption hypothesized that religiosity would be negatively associated with that younger individuals are more likely to trust in science (H1). trust in science (H3). The results were in line with our Although our finding is supported by some of the previous assumption, however when in addition to all other research [7], it is still somewhat surprising, since general trust is sociodemographic variables, social media use was inserted into known to increase with age [19]. Negative relationship between the model, religiosity was no longer a significant predictor of trust in science and age could be explained by lower average trust in science. Indeed, an unusual positive correlation could be educational levels among the elderly, as both knowledge about observed between religiosity and social media use as an science and certain cognitive skills, which are thought to be information source about COVID-19 ( r = .21, p < .001). A related to higher trust in science [6][20] are developed through possible explanation for this phenomenon may be that social education [16][21]. Indeed, in our study age and education were media use is highly prevalent among religious individuals since negatively correlated ( r = -.12, p = .010), however age remained social networks are often seen as channels that can be used to a significant predictor even when education was controlled for. effectively minister to others [25]. Obtaining (mis)information Another possibility may be that the relationship between age and from social media may thus be a side effect of extensive use of trust in science is underlaid by religiosity, as previous research social networks for other purposes. An alternative explanation showed that older individuals are more likely to be religious [22] may also be that religious individuals are more likely to adopt and that religiosity also predicts lower trust in science [10]. conspiracy beliefs [15]. Since conspiracy ideation is likely to However, our results showed that religiosity and age were not influence the perception of traditional media as deceiving [26], significantly correlated ( r = -.03, p = .504), therefore those who are more religious may thus be inclined to use undermining the described reasoning. informal sources of information, such as social media. The results of our study were also in line with the assumption Regarding education and social media use, the results that male gender would be positively related to trust in science supported both of our hypotheses that trust in science would be (H2). Even though some of the previous studies [8] indicated that positively related to years of education (H4) and negatively this relationship could be entirely accounted for by other related to perceived share of information about COVID-19 that sociodemographic variables, we found that gender remained a was obtained on social media (H5). Although more educated 39 Information Society 2022, 10-14 October, Ljubljana, Slovenia Ž. Zelič et al. individuals were also less religious ( r = -.10, p = .023) and [4] Florian Wintterlin, Friederike Hendriks, Niels Mede, Rainer Bromme, obtained smaller share of information on social media ( r = -21, p Julia Metag and Mike Schäfer, 2022. Predicting public trust in science: The role of basic orientations toward science, perceived trustworthiness of < .001), education remained an important predictor of trust in scientists, and experiences with science. Frontiers in Communication 6 science even when other variables were controlled. As (Jan, 2022). DOI: 10.3389/fcomm.2021.822757 [5] Louis Nadelson, Cheryl Jorcyk, Dazhi Yang, Mary Jarratt Smith, Sam previously suggested, this could be explained by the fact that Matson, Ken Cornell and Virginia Husting, 2014. Trust in science and critical thinking, which is thought to interrelate with trust in scientists. School Science and Mathematics 114 (Jan, 2014), 76–86. DOI: science [20], develops through education [16]. Furthermore, in 10.1111/ssm.12051 [6] Fabienne Crettaz von Roten, 2004. Gender differences in attitudes toward contrast to previous research that reported on the positive science in Switzerland. Public Understanding of Science, 13 (Apr, 2004) relationship between social media use and trust in science [13], 191–199. DOI: 10.1177/0963662504043870 [7] Paul Brewer and Barbara Ley, 2013. Whose science do you believe? our results showed that in the times of the COVID-19 pandemic Explaining trust in sources of scientific information about the obtaining information from social media might in fact be environment. Science Communication 35, 1 (Feb, 2013), 115–137. DOI: 10.1177/1075547012441691 detrimental for trust in science. Since social media’s regulations [8] Bernadette C. Hayes and Vicki N. Tariq, 2000. Gender differences in on shared content are less strict compared to the traditional media, scientific knowledge and attitudes toward science: a comparative study of we believe the quick dispersion of COVID-19 conspiracy beliefs four Anglo-American nations. Public Understanding of Science 9 (Oct, 2000), 433–447. DOI: 10.1088/0963-6625/9/4/306 through social media could lower trust in science. Additionally, [9] John Evans and Michael Evans, 2008. Religion and science: Beyond the we hypothesized that social media use would not reflect in high epistemological conflict narrative. Annual Review of Sociology 34 (Aug, 2008), 87–105. DOI: 10.1146/annurev.soc.34.040507.134702 levels of scientific distrust as long as the individuals would be [10] Jonathon McPhetres and Miron Zuckerman, 2018. Religiosity predicts sufficiently educated (H6). We assumed that well educated negative attitudes towards science and lower levels of science literacy. individuals would be able to critically evaluate the quality of PLoS ONE 13, 11 (Nov, 2018), Article e0207125. DOI: 10.1371/journal.pone.0207125 obtained information due to their advanced critical thinking skills [11] Amitai Etzioni and Clyde Nunn, 1974. The public appreciation of science [16]. In contrast to our expectations, the results showed that there in contemporary America. Daedalus 103, 3 (Jan, 1974), 191–205. DOI: 10.1007/978-94-010-1887-6_15 was no significant interaction between education and social [12] Hee-Je Bak, 2001. Education and public attitudes toward science: media use when predicting trust in science. In our opinion, this Implications for the “Deficit Model” of education and support for science and technology. Social Science Quarterly 82, 4 (Dec, 2001), 779–795. finding could be based on the fact that: a) years of education are DOI: 10.1111/0038-4941.00059 not a valid indicator of critical thinking skills, or b) that critical [13] Brigitte Huber, Matthew Barnidge, Homero Gil de Zúñiga and James Liu, thinking abilities are somewhat irrelevant in the case when one’s 2019. Fostering public trust in science: The role of social media. Public understanding of science 28, 7 (Sep, 2019), 759–777. DOI: information space is so limited that they do not have any relevant 10.1177/0963662519869097 data upon which information from social media could be judged. [14] Ramez Kouzy, Joseph Abi Jaoude, Atif Kraitem, Molly B. El Alam, Basil Karam, Elio Adib, Jabra Zarka et al., 2020. Coronavirus goes viral: To conclude, our findings suggest that in order to restore trust Quantifying the COVID-19 misinformation epidemic on Twitter. Cureus in science and reinforce health-related behavior in the context of 12, 3 (Mar, 2020), Article e7255. DOI: 10.7759/cureus.7255 the pandemic, it would be expedient to develop communication [15] Žan Zelič, Martin Berič and Darja Kobal Grum, 2022. Examining the role of Covid-19 conspiracy beliefs in predicting vaccination intentions, strategies that would specifically target older women, who are preventive behavior and willingness to share opinions about the less educated, more religious and are extensive social media coronavirus. Studia Psychologica 64, 1 (Mar 2022), 136–153. DOI: 10.31577/sp.2022.01.844 users. However, these findings are subjected to some limitations [16] Christopher R. Huber and Nathan R. Kuncel, 2016. Does college teach of our research design. Firstly, the data may not be entirely critical thinking? A meta-analysis. Review of Educational Research 86, 2 representative due to the convenience sampling method. (Jun, 2016), 431–468. DOI: 10.3102/0034654315605917 [17] Yves Rosseel, 2012. lavaan: An R package for structural equation Secondly, correlational design of our study does not allow for modeling. Journal of Statistical Software 48, 2 (May, 2012), 1–36. DOI: causal inferences. And thirdly, the used trust in science measure 10.18637/jss.v048.i02 [18] IBM Corp. 2017. IBM SPSS Statistics for Windows, Version 25.0. IBM was one-dimensional, although some researchers argue that it is Corp., Armonk, NY. necessary to differentiate between trust in scientific methods and [19] Michael Poulin and Claudia Haase, 2015. Growing to trust: Evidence that trust increases and sustains well-being across the life span. Social trust in scientific institutions [27]. Therefore, our suggestion for Psychological and Personality Science 6, 6 (Mar, 2015), 614–621. DOI: future research would be to examine how these two distinct 10.1177/1948550615574301 forms of trust in science relate to health behavior and to identify [20] John Kleinig, 2016. Trust and critical thinking. Educational Philosophy and Theory 50, 2 (Jun, 2016), 1–11. DOI: which are the most important risk factors for either of them. 10.1080/00131857.2016.1144167 [21] Bryan Kennedy and Meg Hefferon, 2019. What Americans know about science? Pew Research Center, Washington, DC. Available at: ACKNOWLEDGMENTS https://www.pewresearch.org/science/2019/03/28/what-americans-know-Research was supported by the Slovenian research agency [grant about-science/ [22] Vern L. Bengtson, Merril Silverstein, Norella M. Putney and Susan C. number P5-0110]. Harris, 2015. Does religiousness increase with age? Age changes and generational differences over 35 years. Journal for the Scientific Study of Religion 54, 2 (Sep, 2015), 363–379. DOI: 10.1111/jssr.12183 REFERENCES [23] Global Gender Gap Report 2022. World Economic Forum, Geneva. [1] Office for National Statistics, 2021. Leaving no one behind – a review of Available at: https://www.weforum.org/reports/global-gender-gap-report-who has been most affected by the coronavirus pandemic in the UK: 2022/ December 2021. Office for National Statistics, Newport. Available at: [24] Mary F. Fox and Glenn Firebaugh, 1992. Confidence in science: The https://www.ons.gov.uk/economy/environmentalaccounts/articles/leavin gender gap. Social Science Quarterly 73, 1 (Mar, 1992), 101–113. gnoonebehindareviewofwhohasbeenmostaffectedbythecoronaviruspande [25] Pamela J. Brubaker and Michel M. Haigh, 2017. The religious Facebook micintheuk/december2021#economic-impact experience: Uses and gratifications of faith-based content. Social Media + [2] Yann Algan Eva Davoine, Martial Foucault and Stefanie Stantcheva, Society 3, 2 (Apr, 2017), 1–11. DOI: 10.1177/2056305117703723 2021. Trust in scientists in times of pandemic: Panel evidence from 12 [26] Stephen Marmura, 2014. Likely and Unlikely Stories: Conspiracy countries. Proceedings of the National Academy of Sciences of the United Theories in an Age of Propaganda. International Journal of States of America 118, 40 (Sep, 2021), Article e2108576118. DOI: Communication 8, 1 (May, 2014), 2377–2395. 10.1073/pnas.2108576118 [27] Peter Achterberg, Willem de Koster and Jeroen van der Waal, 2017. A [3] Bernard Barber, 1987. Trust in science. Minerva 25 (Mar, 1987), 123–134 science confidence gap: Education, trust in scientific methods, and trust in DOI: 10.1007/BF01096860 scientific institutions in the United States, 2014. Public understanding of science 26, 6 (Dec, 2015), 704–720. DOI: 10.1177/0963662515617367 40 Indeks avtorjev / Author index Bass-Krueger Julian ....................................................................................................................................................................... 5 Berič Martin ................................................................................................................................................................................. 37 Bratuša Maša ................................................................................................................................................................................ 10 Demšar Ema ................................................................................................................................................................................... 5 Gsenger Rita ................................................................................................................................................................................. 14 Heintz Christophe ......................................................................................................................................................................... 33 Justin Martin ................................................................................................................................................................................. 18 Kobal Grum Darja ........................................................................................................................................................................ 37 Marušič Jar Žiga ........................................................................................................................................................................... 22 Tomat Nastja ................................................................................................................................................................................ 27 Voronina Liubov .......................................................................................................................................................................... 33 Wiedemann Elisa............................................................................................................................................................................ 5 Zelič Žan ...................................................................................................................................................................................... 37 41 42 Kognitivna znanost Cognitive Science Uredniki  Editors: Toma Strle, Borut Trpin, Olga Markic Document Outline 02 - Naslovnica - notranja - B - TEMP 03 - Kolofon - B - TEMP 04 - IS2022 - Predgovor - TEMP 05 - IS2022 - Konferencni odbori - TEMP 07 - Kazalo - B Blank Page 08 - Naslovnica - notranja - B - TEMP Blank Page 09 - Predgovor podkonference - B 10 - Programski odbor podkonference - B 01 - Bass-Krueger-et-al_finalfinal 02 - bratusa_final 03 - Gorjup - PRAZNO 03 - Gorjup - PRAZNO 03 - Gorjup - PRAZNO 03 - Gorjup - PRAZNO 03 - Gorjup - PRAZNO 04 - Gsenger_final 05 - Justin_final Abstract 1 Introduction 2 Black box problem 3 Explanation and understanding 4 Two dimensions of opacity 5 Prediction versus understanding 6 Conclusion Acknowledgments 06 - Marusic_final 07 - Tomat_final 08 - Voronina_final 09 - Zelic_et_al-final 12 - Index - B Blank Page 07 - Kazalo - B.pdf Blank Page 12 - Index - B.pdf Blank Page 08 - Naslovnica - notranja - B.pdf Blank Page Blank Page