ABSTRACT. The two-headed, "fast-and-slow" dual-process view of human thinking has inspired numerous psychologists, economists, and philosophers. However, despite the popularity the framework faces multiple challenges. This symposium will familiarize the audience with recent empirical work and dual process theorizing advances that address critical outstanding issues.

Organizer: Wim De Neys (Université Paris Cité, France)

Speakers and talks:

The role of meta-control in modulating thinking Miroslav Sirota (University of Essex, UK)

Logic or heuristic? Examining the source of intuitive inferences Omid Ghasemi (University of New South Wales, Australia)

System 1 debiasing and logical intuitions Esther Boissin (Cornell University, USA)

Decoding Dual Processes: Insights from Computational Modelling Zoe Purcell (Université Paris Cité, France)

A framework for understanding reasoning errors: From fake news to climate change and beyond Gordon Pennycook (Cornell University, USA)

ABSTRACT. People routinely use samples of information obtained from their social environment or personal experience to make inferences, judgments and decisions. Notably, inferential errors can arise from reliance on unrepresentative or biased samples. This symposium will present state-of-the-art work from five research programs that seek to advance our understanding of the role of information sampling in reasoning and decision-making. The talks will outline recent findings and computational modeling of how sampling processes affect inferences and decisions in individual and group contexts. Gaël Le Mens will examine how the mapping between mental categories and the structure of the environment affects patterns of over- and under-generalization in sampling tasks. Brett Hayes will show how incomplete sampling of inferential hypotheses can produce “learning traps” that are resistant to change. Joakim Sundh will discuss the processes involved in sampling relevant exemplars from long-term memory to make judgments and decisions. Manikya Alister will examine how people adjust their inferences depending on whether they believe information is sampled from a helpful or deceptive social informant. Finally, Gordon Brown will review recent advances in Social-Sampling-Theory which examines how private attitudes and social norms interact to produce phenomena such as polarization and contagion effects in social networks.

ABSTRACT. The endorsement of pseudoscientific beliefs can seriously jeopardize the well-being of patients, so it is necessary to find strategies to tackle it. Causal illusions, the erroneous perception of causal connections between non-contingent events, have been proposed as a cognitive facilitator of these beliefs. Additionally, previous research has demonstrated that the format in which contingency information is displayed can impact causal judgments. We conducted three preregistered experiments to examine the effect of graphical displays on the strength of causal illusions and reasoning strategies. Study I revealed that frequency trees and contingency tables with icon displays lead to weaker causal illusions than trial-by-trial presentations or contingency tables with numbers. In Study II, we manipulated two presentation aspects of contingency tables. Although these manipulations exerted no effect on the strength of causal illusions, qualitative analysis of the participants’ open responses indicated that the use of more sophisticated reasoning strategies renders weaker causal illusions. In Study III, we directly compared frequency trees and contingency table visualizations and corroborated a benefit of the former. Additionally, we found that complex strategies were more frequently used when information was presented in frequency trees. Altogether, our findings suggest that the efficacy of frequency trees in reducing causal illusions may rely on the fact that they make sophisticated strategies more accessible.

ABSTRACT. Human ability to accumulate and transmit cultural information is a main driver of our species' astonishing technological progress: gradual refinement of tools, techniques, and artifacts has enabled the development of increasingly complex material and immaterial innovations. However, the potential benefits of certain mechanisms contributing to cumulative cultural evolution (CCE) are still widely debated: while some studies show that optimized technologies can result from an accumulation of small improvements over time without requiring causal understanding, others argue that technical reasoning (TR), a form of causal reasoning directed toward the physical world, necessarily accompanies the improvement and transmission of technology. Here, we seek to determine the extent to which TR is beneficial for CCE. We ran a transmission chain experiment in which participants were asked to solve one of two versions of a computer-based task. In the Concrete Task treatment, participants faced a simulated version of an actual physical task, which allowed them to rely on technical reasoning to improve their solution. In the Abstract Task treatment, participants faced a task with the same solution-score mapping, but preventing them from relying on TR to improve their solution. This design allows us to investigate two pivotal questions about the role of TR for CCE: (1) To what extent does TR guide individuals’ exploration of the design space?, (2) To what extent does TR lead to a higher pace of CCE?

ABSTRACT. Absence is the negation of presence, and presence is the negation of absence, but they are not cognitively symmetrical. Our attention is naturally drawn to what is present, while what is absent tends to fade away in the background. This asymmetry (i.e., figure-ground relationship; Hattori et al., 2016) serves as one of the contributing factors for anomalies in thinking. In causal cognition, the asymmetry between occurrence and non-occurrence of action or outcome does matter. Instances with non-occurrence tend to be overlooked. This can lead to problems in probabilistic reasoning, as seen in the infamous Monty Hall problem. Its difficulty arises from the challenge of discerning causal structures. By reframing the host's act of "opening the door" as preserving the winning option instead of removing a losing option (i.e., figure-ground reversal), people are expected to gain insights into the task structure. Six experiments were conducted with 1519 participants. The results revealed that making causal relationships clear can increase correct responses, while several other factors also suggested contributing to the problem. This study uncovers the role of figure-ground framing in causation, which can have a significant impact on even apparently uncausal task performances.

ABSTRACT. We built a novel experimental paradigm to explore how causal-selection explanations influence abductive causal inference. According to current theories, humans intuitively assign a causal strength score to each subset of variables that contributed to an outcome, that depends both on a variable’s causal link to the outcome and its prior probability. These scores determine which events subjects consider as the main causes of an outcome, otherwise known as causal-selection judgments. Our paradigm combined observational data with explanations consisting of such judgments in a causal-inference task. Participants observed outcomes by drawing balls from four urns, each with their own proportion of colored balls. Their goal was to guess the rule linking draws to the outcome based on the information provided. One group received causal-selection judgments as explanations for each draw, pointing to the cause that extant theories predict to have the highest causal strength. Another group saw explanations that pointed to an event other than the top-ranked cause, while still an active cause of the outcome. A third group was only given observations with no explanations attached. We provide empirical and computational results showing people are most likely to correctly identify the underlying causal structure when given causal selection explanations.

ABSTRACT. Investigating how people evaluate more or less complex explanations of the same evidence has been a focal point of research. However, previous studies have either focused on the choice between a limited set of explanations or could not systematically quantify the complexity of the explanations participants favoured. We provide a new approach for modelling explanation selection which foregrounds the balance between observed and latent structure in the mechanism being explained (what proportion of its components are observable). We combine a Bayesian framework with program induction, enabling coverage of unbounded partially observable model space through sampling, and reflecting how a simplicity bias emerges naturally in this setting. Through simulation, we identify two novel principles: (1) simpler explanations should be favoured as latent uncertainty (the number of hidden variables in the system being explained) increases and (2) latent structure is attributed to a larger role when the observable patterns become less compressible. We designed a behavioural experiment and found that these principles were reflected in human judgments, indicating that people are sensitive to latent uncertainty when selecting complex versus simple explanations.

ABSTRACT. Behavior can be significantly influenced by limited rationality, often seen as a result of restricted interactive reasoning. However, these limitations might not solely depend on reasoning abilities. In this paper, we try to disentangle ability and beliefs about rationality in an epistemic game theory framework. To test our identification strategy, we report results from an experiment where subjects play two perturbed ring games that differ in whether beliefs are endogenous or exogenously fixed. That is, the subjects play once against other humans and once against a set of computers whose beliefs are controlled. This is meant to allow us to infer the subjects’ ability bounds from their decisions against computers, which allows us to identify their beliefs about the rationality of their human counterparts. In addition to their experimental behavior, we assessed the subject’s IQ with a CFT 20-R. Surprisingly, we find that our subjects perform better against other humans than against computers and that intelligence only predicts performance against computers and not humans. We discuss the results in the context of epistemic game theory as a descriptive theory of decision-making.

ABSTRACT. In this paper we test and compare several message-based nudges designed to promote civil discourse and reduce the circulation of hate speech. We conducted a large pre-registered experiment (N = 4,081) to measure the effectiveness of seven nudges: making descriptive norms, injunctive norms, or personal norms salient, cooling down negative emotions, stimulating deliberation or empathy, and highlighting reputation. We used an online platform that replicates a social media newsfeed and presented the nudge as a message when entering the platform. Our findings indicate that nudges making descriptive norms salient and cooling down negative emotions selectively increase participants' overall engagement with relatively harmless content. Additionally, making injunctive norms salient is effective at increasing post-liking for harmless content. Exploratory text analysis also reveal that highlighting reputation led to more substantial and self-consistent comments on harmful posts. These results suggest that message-based nudges on norms and emotions represent a promising approach to promoting civil discourse and making social media a safer and more inclusive space for all.

ABSTRACT. The present study aims to test the effect of exposure to news in a social media environment on people’s perceived knowledge of selected topics and on the “illusion of knowledge” effect, i.e., the overestimation of one’s perceived knowledge relative to one’s actual knowledge. We furthermore investigate how the effect of exposure varies depending on the level of self-involvement in the topics covered by the news. The research protocol consists of an online study composed of pre-exposure assessment, stimuli presentation, and brief post-exposure questionnaires. The study employs a mixed design, and it is divided into two sessions, scheduled two weeks apart. Participants will be randomly assigned to one of three experimental groups, characterized by the content of the newsfeed they will scroll through. Participants will be asked to assess their perceived knowledge of several topics, before (T1) and after (T2) having scrolled through a mock social media news feed resembling Facebook’s, where they will find news articles about two of those topics. In addition, perceived knowledge will be compared to a standardised test of factual knowledge to measure the possible presence of the illusion of knowledge. We hypothesize that social media exposure will increase participants’ perceived knowledge and that such an increase will be greater for participants exposed to topics perceived as more involving. We further expect participants’ perceived knowledge to be unmatched by their actual knowledge, thus observing illusion of knowledge, and that this phenomenon will be similarly affected by exposure and perceived involvement in the topic. This discrepancy will be tested across groups to check whether it is enhanced by news exposure.

ABSTRACT. Two pre-registered experiments (each 256 participants) investigated repeated choice between a pair of binary-outcome options, with immediate feedback for obtained and forgone outcomes. Outcome-probabilities changed (unannounced) twice across 180 choice-rounds (Experiment 1), or thrice in 160 rounds (Experiment 2). Those changes either preserved or reversed which option was best. We measured and compared the effects on subsequent choices of obtaining disappointment-inducing outcomes (receiving the lower outcome from an option) and regret-inducing outcomes (obtained outcome < forgone outcome). We also generated predicted choice proportions from disappointment-based and regret-based variants of a ‘win-stay-lose-shift’ model, testing both models against a positive-reinforcement model.

The regret-based model described behaviour best. First, the absence/presence of regret-inducing outcomes predicted ‘staying’ or ‘shifting’ more strongly than disappointment-inducing outcomes (d > 1.3). Second, observed choice proportions were typically closer to those predicted by the regret-based model, as compared to both other models. Importantly, the regret-based model best predicted the direction and size of behaviour changes that followed changes in outcome-probabilities. Additionally, individual task performance correlated positively with the proportion of a participant’s choices matching those prescribed by the regret-based model. Our data suggest experienced regret can be an important driver of choice, contributing to simple but adaptive choice strategies.

ABSTRACT. Abstract ICT

Decisions about extinction risk are ubiquitous in everyday life and for our continued existence as a species. We introduce a new risky-choice task that can be used to study this topic: The Extinction Gambling Task. Here, we investigate two versions of this task: a Keep variant, where participants cannot accumulate any more earnings after the extinction event, and a Lose variant, where extinction also wipes out all previous earnings. We derive optimal strategies for both variants and compare them to participants' behaviour in a series of four experiments. Our findings suggest that people understand the difference between the variants and their behaviour is qualitatively in line with the optimal solution. We further test heuristic accounts of the strategies that participants use to approximate the optimal solutions. Finally, we provide mixture modelling results to better understand variability between participants in terms of the employed strategies. We hope that our task and results will motivate further research on this vital topic.

ABSTRACT. This study investigates to what extent the 'green gap,' i.e. the disconnection between environmental values and pro-environmental behavior, can be explained by various perceived psychological barriers, such as conflicting goals, perceiving change in behavior as unnecessary, not having enough information about how to change one’s behavior, or feeling that one is already doing enough. A large representative sample of Slovak participants (N = 1233) filled in several measures of environmental value orientation, perceived psychological barriers and various self-reported pro-environmental behaviors (e.g. transportation, diet adjustment, social behavior, buying seasonal, local, and packaging-free products, and reducing energy and water use, and waste). The parallel mediation models revealed that environmental value orientation predicts engagement in pro-environmental behavior, and this relationship is partially mediated by certain psychological barriers, yet, the role of the specific barriers varied substantially with different types of behavior. Overall, however, the explained variance in pro-environmental behavior was relatively modest (up to 31%), suggesting the presence of other important social, psychological, as well as structural variables as predictors in this regard. Our findings emphasize the intricate role of psychological barriers and demographic factors in shaping environmental actions. Understanding these complexities can help design targeted interventions to bridge the 'green gap' effectively.

ABSTRACT. In the present study, we examined the role of cognitive factors as predictors of psychological barriers to pro-environmental behavior and self-reported pro-environmental behavior with attitudes and beliefs toward climate change as mediators of this relationship. A large representative sample of Slovak participants (N = 1233) filled in several measures of cognitive predictors (education, scientific reasoning skills, conspiracy belief, and distrust in science), attitudes and beliefs towards climate change (climate skepticism and psychological distance), perceived psychological barriers (various psychological reasons why people do not adopt pro-environmental behavior even if they accept the severity of the issue of climate change) and self-reported pro-environmental behaviors. Our analyses show that while our mediation model explains 29% of the variance in psychological barriers, it only explains 4% of variance in pro-environmental behavior. The most important mediator in both relationships was climate skepticism. These findings suggest that while cognitive factors are not among the most important predictors of pro-environmental behavior, they substantially contribute to various antecedents of it, such as beliefs and attitudes and perceived psychological barriers with regard to climate change.

ABSTRACT. This longitudinal study investigates how decision-making styles (rational and avoidant) influence students' self-efficacy and subsequently, their goal progress and action crisis. The sample consists of 108 Slovak grammar school students (64.8% female, mean age 18.5) whose goal is to be accepted to university. The first data collection included the GDMS questionnaire and three goal self-efficacy questions. Goal progress and action crisis were assessed in the third data collection. The full mediation models revealed a significant mediating role of self-efficacy between both rational (indirect effect b = 0.59, p = 0.09, 95% CI [0.05, 1.45]) and avoidant (indirect effect b = -0.59, p = 0.01, 95% CI [-0.19, -2.59]) decision-making styles and progress. The mediating effect of self-efficacy between the rational decision-making style and action crisis was not significant (total effect b = -0.35, p = 0.16, 95% CI [-0.73, 0.00]) although it was in the case of the avoidant decision-making style (total effect b = 0.44, p < .001, 95% CI [0.22, 0.65]). The results demonstrate the negative effect of avoidant decision-making on self-efficacy, subsequently hindering progress and leading to the development of action crises. However, rational decision-making was found to support self-efficacy, thereby facilitating progress towards our goals.

ABSTRACT. Background The two recent studies (Adam-Troian et al., 2023; Adamus et al., 2024) corroborated the view that institutional trust mediates the relationship between subjective appraisal of financial insecurity and conspiracy beliefs. While both studies used large, international samples, and delivered interesting findings that contributed to our knowledge about structural antecedents of conspiracy beliefs, the joint limitation of the two studies is that the data used were mostly cross-sectional. Consequently, the studies refrained from drawing causal about the associations between those three variables. It is an important drawback because the literature argues that the relation between people’s distrust in political institutions and their adherence to conspiracy beleifs is in all likelihood bidirectional (van Prooijen et al., 2022). Moreover, the past research shows that institutional trust is sensitive to contextual changes (e.g., elections, economic crises) and could reflect changes of confidence in institutions during particularly turbulent times (Hornsey et al., 2023).

Objectives Thus, the present study directly responses to the call by the authors (Adam-Troian et al., 2023; Adamus et al., 2024) to perform longitudinal and/or experimental studies to delve deeper into possible causal relations between the financial insecurity, institutional trust and the endorsement of conspiracy beliefs. To meet this aim, the current research attempts to establish how the patterns of relationships last (or change) throughout time. The study aims at a secondary analysis of the available longitudinal data collected during a three-wave mini-longitudinal study in Slovakia. The datasets analysed in this study are part of a larger project and have already been analysed for other research objectives. Data from the first two waves of the study were also analysed within the Adamus et al. (2024) study that replicated the mediation model postulated by Adam-Troian et al. (2023).

Methods Financial insecurity. Subjective appraisal of financial insecutiry was measured using six items that asked participants to indicate how they feel about their current financial situation (e.g. “How uncertain do you feel?”) from the modified version of the Financial Threat Scale (FTS) (Marjanovic et al., 2015). Participants were required to indicate how they rated the stability and security of their personal finances during the pandemic (5-point scale, from 1 = not at all to 5 = very much).

Conspiracy beliefs. Conspiracy beliefs were measured using 18 items from the COVID-19 Unfounded Beliefs Scale (C19-UB) (Teličák & Halama, 2022). Participants were asked to indicate their agreement with conspiracy beliefs related to COVID-19 (5 items, for example, “COVID-19 was planned long ago to weaken the economy and cause unemployment”), measurements (4 items, for example, wearing protective masks is dangerous for schoolchildren and the elderly), treatment (5 items, for example, intravenous (injectable) use of sodium chlorite has proven to be effective against COVID-19) and vaccinations (3 items, for example, vaccines against COVID-19 contain substances that cause infertility or abortion). Higher mean scores indicate higher endorsement of conspiracy theories (5-point scale, from 1 = totally disagree to 5 = totally agree).

Trust. Institutional trust was measured as institutional trust. We asked participants to what extent they personally trusted each of 8 institutions, or 10 institutions (for example, Ministry of Health of the Slovak Republic, European Medicines Agency, Slovak Academy of Sciences, doctors and health professionals) in the second wave (10-point scale, from 1 = absolutely do not trust to 10 = absolutely trust).

Analytical approach We will test a three-wave autoregressive longitudinal mediation model with precarity included as a predictor (X), institutional trust as a mediator (M) and conspiracy beliefs as outcome (Y). Crucially, the relations between X, M and Y will be allowed to vary freely (not only presumed mediation paths but also opposite paths will be estimated). Estimating all paths and then testing the fit of the constrained models will allow us to test whether our presumed longitudinal association (precarity > institutional trust > conspiracy beliefs) fits the data better than the opposite longitudinal associations.

ABSTRACT. In 2023, Suri et al asked chatGPT 3.5 “Do you think you, ChatGPT, will show decision-making heuristics and biases such as anchoring?” and the answer was that in its role as an artificial intelligence language model, the system lacked the ability to demonstrate cognitive biases such as the anchoring heuristic, commonly found in human thinking and decision-making. The model also discussed how its responses are crafted from patterns identified in vast language datasets, emphasising that these responses may inadvertently mirror biases or heuristics present in the training data, potentially perpetuating them in the generated content. Only a few months after, 50 iterations of this exact same query in chatGPT3.5 provided very different results. In 46 out of 50 iterations (92%), ChatGPT 3.5 explicitly said yes. In the remaining 4, ChatGPT 3.5 denied possessing personal beliefs or cognitive processes and not engaging in decision-making heuristics and biases in the same way humans do and having personal thoughts, emotions, or biases or possessing personal opinions, thoughts, or intentions. To further explore these answers, as well as investigate different methodologies for studying generative AI, we will involve ChatGPT 3.5 in a semi-structured interview aimed at exploring its levels of meta-cognition and bias awareness.

ABSTRACT. Shifting from artificial to natural materials is recognized as a prominent strategy to mitigate the environmental crisis. Research has shown a preference for natural materials in safety- and health-focused domains, the so-called naturalness bias. However, little is known about performance-driven domains, where market data suggest a generalized preference for artificial materials. This research investigates the existence of an artificiality bias in performance-driven domains, i.e. a systematic tendency to overvalue artificial solutions and undervalue natural ones. We suggest that a belief in human dominance over nature underlies this bias, leading individuals to perceive human-made alternatives as superior. We identify information provision as a potential debiasing strategy. A pilot study (N=125) affirmed the belief in human dominance over nature, paving the way to testing its effect on consumers’ preferences. In a lab experiment (N=188), it was observed that consumers exhibit a growing inclination towards artificial solutions when they prioritize performance as a key product attribute. Furthermore, an online experiment (N=428) indicated that, in such scenarios, preferences for natural options emerge only when individuals receive reassuring information about their performance. These findings hold significance for policymakers, marketers, and environmental advocates aiming to promote the adoption of natural materials for a sustainable future.

ABSTRACT. Numerous studies have investigated verbalization effects with the aim of clarifying the processes responsible for solving insight and non-insight problems. However, there is no concordance among the various theories in the literature. While studies supporting the special process view have reported different effects of verbalization in insight and non-insight problems, highlighting an overshadowing effect in the former, studies accounting for the business-as-usual approach reported a lack of verbalization effect in insight problems and claimed that the solution processes of the two types of problems are more similar than different. The study in the present paper investigates, with the same procedure, the effect of verbalization on insight and non-insight problems, providing evidence supporting the role of unconscious processes for the resolution of the former, in contrast with the stepwise, conscious procedure for the resolution of the latter. Results will be discussed in the light of a third approach, Unconscious Analytic Thought (Macchi and Bagassi, 2012, 2015; Bagassi and Macchi, 2016), which claims that the insight problem-solving process is mainly unconscious and analytic.

ABSTRACT. Recent research found that differences in children’s performance on Raven’s Standard Progressive Matrices (SPM), one of the most widely used reasoning tests, are associated to differences in the way children adaptively modulate their study times as a function of item difficulty (Perret & Dauvier, 2018). However, the reason why older children allocate more time to difficult items must be clarified. This may reflect developmental differences in strategy adaptivity (Gonthier, Harma & Gavornikova-Baligand, 2023), differences in motivation (which appears to be related to adults’ RT on difficult items: Gonthier & Roulin, 2020), or differences in metacognitive sensitivity to difficulty changes. The present study investigated these three hypotheses with 165 children aged 7 to 11 years, who completed a digitized short form of the SPM. The task used a split-screen method to assess strategy use (Rivollier et al., 2021), and children also estimated the difficulty of each item and completed two motivations questionnaires. The results showed that both the sensitivity of difficulty judgments and strategy use contribute to the adaptive modulation of study times. These results broadly support the significant role of monitoring-control relations in reasoning efficiency (Ackerman & Thompson, 2017) and shed light on age-related differences in meta-reasoning processes.

ABSTRACT. Previous research has studied age differences in affective decision making through the observation of behavioral trajectories in the Iowa Gambling Task (IGT) over the course of the life span. While studies on adolescents have used the original version of the IGT, a few studies have tested a simplified 2-deck version on 3 to 7 years old children. To our knowledge, the IGT has not yet been adapted for preadolescents. Because of this gap, we created a computerized, 4-deck version of the IGT specifically designed for preadolescents, namely the Children’s Gambling Tasks (CGT). CGT involves executive functions (EF), which are responsible for initiation, planning, organization, and regulation of behavior. Since EF – particularly inhibition, working memory, and cognitive flexibility – are predictors of decision-making performance, as confirmed by numerous empirical findings, the relationships between CGT and EF are worth investigating in preadolescence. To explore the predictive role of EF on affective decision making, sixty-seven healthy children aged 8-11 solved the CGT and completed an assessment of inhibition, reasoning, working memory, and divergent thinking. Results revealed diverse predictions of EF on CGT overall performance (all blocks), decision making under ambiguity (blocks 1 and 2), and decision making under risk (blocks 3, 4, 5). These findings might expand the understanding on how cognitive development affects decision-making processes in preadolescence and may have implications for future research on developmental trajectories in CGT measures.

ABSTRACT. Recent studies examined how psychological inflexibility, from a cognitive and social perspective, can affect individual functioning and decision-making in everyday life. In this regard, the construct of socio-cognitive polarization (SCP) has been introduced, describing the tendency to adopt extreme opinions as a cluster measure of conservative political tendencies, absolutistic thinking, and xenophobic attitudes. Among the psychological biases that can affect individual choices, the status quo bias (SQB) – namely the preference to maintain and favor current or already existing states of the world – may lead people to make irrational and ineffective choices when dealing with social and technological matters. The present study aims to explore the relationship between the tendency towards the status quo, differences in SCP levels, and socio-demographic characteristics. One hundred fifty-five Italian adults completed an online survey including the SCP scale and a SQB tendency task. The SQB task consisted of the presentation of 4 hypothetical, albeit plausible, scenarios focused on a specific topic of current relevance to the present social, cultural, economic, and political context. In particular, the scenario referred to possible changes about a) technological progress (sustainability, artificial intelligence) or b) social progress (multi-ethnic integration, gender identity). We aimed to assess whether participants would stick to the default option or if they were inclined to foster the alternative, which promoted a change towards the opposite direction. Results showed that while individuals with higher levels of SCP showed an SQB tendency when judging social changes, they were compliant with technological changes. By contrast, the opposite trend emerged in individuals with the lowest SCP levels. These findings highlighted an interaction effect of status quo tendency and psychological inflexibility on decision-making regarding real-life collective changes.

ABSTRACT. Proliferation of misinformation online constitutes a threat to individuals and society. Adolescents, due to unique cognitive and socio-affective characteristics, could be particularly vulnerable to misinformation. Designing interventions to improve their ability to judge information online has thus become critical. Drawing from two theoretical frameworks (i.e., the dual process theory-DPT and inoculation theory), we designed and compared the effectiveness of two interventions aiming at increasing adolescents’ ability to distinguish between reliable and unreliable content. In the DPT-based intervention, adolescents were trained to overcome intuitive processes and rely on deliberative ones when judging information. In the inoculation-based intervention, adolescents were trained to detect manipulative techniques used to fabricate false information. The study was designed as a large-scale online citizen collaborative research project. Data were collected in 6th to 9th graders, one week before, one week after and 7 weeks after the interventions. Participants completed tasks and scales measuring: 1- their ability to distinguish between real and fake news; 2- their ability to discern between manipulative and neutral information; 3- their sensitivity to conspiracy ideation; 4- their ability to overcome reasoning biases. This study proposes two novel approaches to equip adolescents with the cognitive mindset needed to resist misinformation.

ABSTRACT. Nowadays, it’s common to perform duties in the presence of others (e.g., open landscape offices). However, how the cognitive effort exerted by others may affect our performance remains unclear. Is cognitive effort contagious? Desender et al. (2015) examined this with a modified Simon task by pairing two participants in front of a single yet split screen, where one participant performed a neutral Simon task while their neighbor performed either an easy (mostly congruent block) or a hard one (mostly incongruent block). This aimed to examine whether the neighbor’s exertion of effort would affect the neutral participant’s performance. Indeed, their results corroborated that effort is contagious. In this experiment, we aimed to replicate this result by asking 88 dyads of participants (N = 176) to perform a Simon task while seated next to each other. Like Desender et al. (2015), we manipulated the difficulty of the Simon task that each participant of a dyad saw on their side of the screen. Our study failed to replicate Desender et al.’s results. Unexpectedly, we observed that while the neighboring participants were exerting effort, the neutral ones diminished theirs, showing an opposite result to what we expected based on Desender's findings.

ABSTRACT. The theory of adversarial cooperation holds that a self-organized division of cognitive labor based on complementary, individual biases is collectively optimal for epistemic progress. We report three experiments that support this thesis. Using politicized versions of the Wason Selection Task and the Reduced Array Selection Task, we show that both Democrats and Republicans are more likely to propose empirical tests that can only result in confirming evidence when they have to evaluate a public policy they favor, and more likely to propose empirical tests that can result in disconfirming evidence when they have to evaluate a public policy they disfavor, than when they have to evaluate a neutral policy in the formally identical way. In other words, the space of relevant evidence is collectively better covered in the presence of individual biases than individual impartiality. We also show that the political adversaries of the policies are not just more likely to propose the normatively correct solution on the tasks than political allies, but more likely to provide the correct reasoning for their choice. Lastly, we report findings that reveal whether such insight is gained before or after the adversaries are asked to explain their choices.

ABSTRACT. In order to generate new ideas or solve problems in novel ways, it is important to focus on obscure aspects of things. Yamakawa and Kiyokawa (2020) found that searching for commonalities between apparently unrelated objects encourages attention to obscure features of objects. In order to apply this method to wider problem solving, it is necessary to examine whether this method can be applied to states, events, and concepts other than artefacts. We investigated the effect of commonality search on the perception of 'solitude.' A total of 183 native Japanese speakers participated in an online experiment. The pre-intervention session asked them to provide as many associations with 'solitude' as possible in 30 seconds. Subsequently, their images of 'solitude' were measured using the 11-item SD method. Participants were assigned to a commonality search or reconsideration condition in the intervention session. The participants in the commonality search condition were asked to provide the commonalities between 'solitude' and the words 'winning,' 'blessing,' 'exhilarating,' and 'lucky.' In the reconsideration condition, participants were asked to provide what they associate with 'solitude' four times. The same procedure as the pre-intervention session was used in the post-intervention session. A two-way mixed ANOVA showed a main effect of time on the positive items, indicating a change in the positive direction after the intervention.

ABSTRACT. One of the most investigated properties of causal reasoning is the inference called “explaining away,” which can be observed in a common effect structure involving two independent, non-mutually exclusive causes of the effect. Explaining away means that if the effect is present together with one of the causes, the probability of the other cause is reduced. Magicians often adopt a strategy similar to explaining away by proposing a fake cause (from generic magical abilities/sleight of hand to the power of the mind) to hide the actual cause. In this study, participants watched a magician’s performance video (either a card or a mental magic effect) and were then asked to rate on a Likert scale how easy it was to think of many alternative explanations to understand how this effect works. Each video could come in two versions (randomly assigned to each participant assigned to a specific effect): the performance included a verbal mention of the fake cause or it was verbally omitted. We found the explaining away effect only in the mental magic video. Results suggest that explaining away may work even if the alternative cause is actually unfeasible to achieve the effect and is only verbally mentioned.

ABSTRACT. The foundation of the dual-process theory of thought lies in the coexistence of two distinct modes of thinking: a rapid, automatic, and associative process contrasted with a slow, reflective, and deliberative process. Investigating the neural basis of this dual-process distinction has proven challenging, particularly when applying the standard fMRI approach to tasks commonly used in cognitive literature, leading to mixed results. In our study, we conducted an activation likelihood estimation (ALE) meta-analysis to explore the neural underpinnings of the dual-process theory of thought. The eligible studies allowed us to identify brain regions associated with tasks based on the dual-process theory. The ALE algorithm highlighted activation in the medial frontal cortex, superior frontal cortex, anterior cingulate cortex, insula, and left inferior frontal gyrus. These structures partially overlap with brain areas frequently mentioned in literature discussing the neural basis of the dual-process distinction. The results affirm the potential, yet largely unexplored, common ground between the dual-process literature and the cognitive control literature.

ABSTRACT. A commonly accepted representational change theory links insightful solutions to the mechanism of representational change (Danek, 2018; Ohlsson, 1992; Öllinger et al., 2014). Stellan Ohlsson (2011) argues that representational change occurs as a result of the accumulation of negative feedback on previous unsuccessful actions. We conducted two experiments to examine the role of feedback on behavioural and subjective measures of insight problem solving using compound remote associate (CRA) problems. In both experiments, participants were confronted with CRA problems and had to type in any ideas that they had during the solving process. After submitting each idea, the feedback occurred (either positive, negative, or neutral). In Experiment 1 (N = 35), the feedback came in the form of an increase in a progress bar, while in Experiment 2 (N = 40) in the form of the percentage of people who successfully solved problems after having the same idea. In line with the representational change theory, we expected that the negative feedback would elicit more correct solutions, faster solutions, and more subjectively insightful solutions. Experiment 1 revealed a positive effect of the negative feedback on solution accuracy. In Experiment 2, solutions were faster in the positive feedback condition. No effect on insightfulness ratings was found.

ABSTRACT. Morris and Peng (1994) found that Westerners tend to attribute causality to dispositional factors, while Easterners tend to focus on situational factors when explaining social events. Arai and Yama (ICT 2016) showed that the Japanese tend to focus on dispositional factors rather than situational factors when making causal inductions. The issue of self-responsibility has been discussed in recent years, for example by Sandel (2007, 2020), and Shibanai (2021) examined Japanese beliefs about self-responsibility using data from the fifth wave of the Asian Barometer Survey. The results showed that men are more likely to hold self-responsibility beliefs than women. The purpose of this study is to investigate this phenomenon using a causal induction paradigm. Japanese participants were asked to make causal judgements between fictitious cases of prisoners' crimes (murder or non-murder) and their motives (situational or dispositional). Each case was presented one by one, and the contingency between the crimes and motives was manipulated. The results showed no significant gender differences, and women were not less likely than men to focus on dispositional factors. It is suggested that this discrepancy is due to methodological differences.

ABSTRACT. In a previous study, it was shown that individuals who are more likely to engage in Jumping to Conclusion (JtC), the tendency to make decisions based on little evidence, are more likely to exhibit causal illusions in a contingency learning task. In the present study, we investigated the relationship between JtC and causal illusion by examining whether individuals with higher JtC tendencies (1) were more likely to predict strong causal relationships prior to contingency learning, and (2) had greater confidence in their prior predictions and causal judgments. Results showed that causal illusion was greater among those with higher JtC who made stronger causal judgments prior to contingency learning, suggesting that those with higher JtC tendencies may have difficulty updating their beliefs with evidence.

ABSTRACT. One of the important problems of the cognitive psychology is the relations between emotions and cognitive processes. Isen and colleagues demonstrated that positive emotions lead to increased creativity (Isen,Daubman,Nowicki,1987). We assume that positive emotions can influence the weakening of the Einstellung effect. The main hypotheses: 1. Positive emotions reduce the solution time of the extinction problem. 2. Positive emotions reduce the subjective Aha! experience of the extinction problem. The adapted series of water jar problems (Luchins,1942) was used as a stimulus material: 8 problems are solved according to the same scheme (Einstellung problems) and 1 critical problem is solved according to the different scheme (extinction problem). To induce emotions, videos were shown to the participants. The participants were randomly divided into three experimental groups: A. positive emotions, B. negative, and B. neutral. The Danek and Wiley (Danek,Wiley,2017) questionnaire was used to "Aha!" experience assessment. Participants were 51 people. Results: the solution time of the extinction problems was different in the three experimental groups (F(16,225)=2.199, p=.006). There were no significant differences in "Aha!" experience assessment of the extinction problems in the three conditions. This work was supported by the Russian Science Foundation, project number 22-18-00358, https://rscf.ru/en/project/22-18-00358/.

ABSTRACT. In a world where uncertainty is inherent, humans must navigate the trade-off between exploitation—i.e., choosing the best options given the known states of the world—and exploration—i.e., choosing options that reveal novel information about the hidden (uncertain) states of the world. Although the basic need to reduce uncertainty leads humans to gravitate towards exploitation, uncertainty opens a plethora of possibilities allowing creative thoughts and actions to emerge, which renders exploration an adaptive strategy. It is unclear whether people with a highly creative mind generally tend to actively seek uncertainty when the choice architecture in their environment affords it. Currently, we explore people’s behaviour under different forms of uncertainty and whether this is linked to their ability to generate creative ideas. To test this, we use a multi-armed bandit task, whereby participants (N = 91) repeatedly choose between reward options that offer various forms of uncertainty, namely relative uncertainty (RU), which captures directed exploration, and total uncertainty (TU), which assesses random exploration. Participants’ creativity is captured by an idea generation task with ratings of fluency, flexibility, originality, elaboration, and novelty-and-usefulness-based creativity. Using a mixed-effects model, we test whether individuals’ creativity profiles predict their behavioural signature of exploration-exploitation when making decisions under uncertainty.

ABSTRACT. Decision-making often involves a trade-off between risks and rewards, for which humans are susceptible to biases. Cognitive training could facilitate risk-reward trade-offs, but, for applicability outside the lab, temporal persistence is a potential issue. In this six-month double-blind randomized control study, participants were split into treatment and control groups to complete a decision-making task involving standard monetary gambles. In the initial phase, all participants completed pre-training (Day 1), training (Day 2-7), and post-training sessions (Day 8). During training, one group was given feedback to promote risk-neutral choice (treatment), whereas the other merely practiced the task (control). Following training, choices in the treatment group were significantly more risk-neutral than in the control group (who showed no improvements). Thus, training was necessary and sufficient to induce behavioral changes. The initial phase was followed by repeat assessments at 1, 2, and 6 months. The pattern observed immediately after training was replicated across all follow-up assessments. Thus, cognitive training effects persist for at least 6 months without any top-up training. Computational modeling revealed a complex pattern of change in the feedback group – whereby participants' initial risk preferences partially determined the effect of training on their post-training preferences.

ABSTRACT. When disseminating health information, expert advice often takes precedence over direct evidence. Little is known, however, about how people integrate expert advice and which factors impact it. Across two experiments, we incentivised participants (N = 284) to accurately assess treatment efficacy, and measured how participants integrated expert advice. We found that participants did not optimally integrate expert advice; taking on expert advice less than the optimal standard. Furthermore, how well participants integrated expert advice was impacted by conflicts in information, the quality of information, and the order in which it was presented. A descriptive Bayesian framework was employed to identify potential mechanisms that may underlie suboptimal information integration (i.e., sampling vs. capacity limits). Our findings provide insights into how best to tailor the communication of scientific evidence to the public.

ABSTRACT. It seems the ability to generate persuasive bullshit may be evolutionarily adaptive, as individuals possessing this skill are not only perceived as more intelligent by others but also exhibit higher verbal intelligence (Turpin et al., 2021). On the other hand, people reporting more bullshitting behavior scored lower in intelligence (Littrell et al., 2021). In this study we examined relationship between people's ability to produce convincing bullshit and extent to which they engage in bullshitting. Our preliminary results on a sample of 433 participants suggest that people more capable of producing convincing bullshit are, indeed, perceived as more intelligent (r(359)=.90, p<.001), but ability to produce more convincing bullshit is unrelated to both persuasive (r(433)=-.09, p=.06) and evasive bullshitting (r(433)=.03, p=.49). On the other hand, participants with better bullshit ability were less likely to fall for bullshit. Overall, our results suggest that 'better bullshiter' may not necessarily be a 'bigger bullshiter'.

ABSTRACT. This paper investigates how semantic control differs between high and low creative people. Semantic control (SC) is the ability to flexibly access and manipulate meaningful information; it helps to focus on meanings of a concept that are relevant to a particular context or task, enhances activation of less dominant meanings, and inhibits more frequent but irrelevant synonyms. The quality of SC is conditioned on the ability to flexibly switch and resolve ambiguous tasks (Jackson, 2021). A more flexible control will allow multiple meanings of a particular concept in different environments, allowing the use of a non-frequent concept in a familiar context. Hypothesis: We hypothesized that highly creative people will have more flexible semantic control, because choosing relevant solutions for specific tasks often involves rejecting more familiar and dominant options and finding underutilized alternatives, which is the responsibility of semantic control. Results: the result can be interpreted in two ways: differences between creative and non-creative subjects do exist, but manifested in more "controversial" sentences, qualitative differences in the evaluation of sentences were found, but there were no differences in reaction time

ABSTRACT. Human perception, memory and decision-making are impacted by tens of cognitive biases and heuristics that influence our actions and decisions. Despite the pervasiveness of such biases in human decision making, they are generally neither considered nor leveraged by today's Artificial Intelligence (AI) systems that model human behavior and interact with humans.

A widespread yet understudied cognitive bias in this context is the attractiveness halo effect. According to this bias, perceived attractiveness impacts our judgments of unrelated attributes, such as intelligence, trustworthiness or sociability. Leveraging AI-based beauty filters, we study the attractiveness halo effect on a diverse set of 462 different individuals both on an “unfiltered” and “filtered or beautified” conditions. By means of a large-scale user study with over 2,700 participants, we find that beautified images receive statistically significantly higher ratings not only of attractiveness but also of other attributes such as intelligence, trustworthiness and sociability. While we see strong evidence of the existence of the attractiveness halo effect both before and after applying beauty filters, we see a significant reduction in the strength of the effect in the beautified set. Thus, we also explore these filters as a potential tool to mitigate the attractiveness halo effect.

ABSTRACT. Judgements about others’ behavior is often made based on the relative rank of that behavior. We investigated this in the new domain of pro-environmental behavior, specifically for the categories of energy and water consumption, food (meat) consumption and transport choice. Using unimodal and bimodal distributions, we experimentally manipulated three fictional individuals’ (common points) rank positions while keeping their absolute frequency of behaviors constant. Consistent with previous literature, participants’ judgements about these people’s pro-environmental behavior differed based on their rank position. Rank effects were not moderated by the perceived importance of others’ behavior, the perceived visibility of the behavior, or the perceived level of control. The results of this experiment are in line with a Decision by Sampling account of judgments of pro-environmental behavior, and set a foundation for future research seeking to conduct behavioral interventions (such as rank-based nudges) within this domain. Prior to this, however, future studies should investigate whether the smaller effect sizes found in this experiment, compared to those seen in previous research, are attributable to methodological differences, or the domain itself.

ABSTRACT. The IPCC recommends that low probability climate change predictions (<50%) are described with negative verbal probabilities. Those phrases attract attention to the target outcome’s non-occurrence, while positive ones attract attention to the outcome occurrence (e.g., it is unlikely vs., there is a chance that it will rain). Negative predictions are typically associated with outcomes larger than expected (e.g., it is unlikely that it will rain 20mm when the models show precipitation ranging from 4 to 18mm). We aimed to replicate that negative verbal probabilities would be associated with a fringe outcome, and to test whether it would indicate a lack of consensus in the possibility expressed. In two experiments (N=298 and 302), we compared participants’ interpretations of two positive (there is a small possibility and a small probability) and two negative phrases (unlikely and the likelihood is low) that conveyed a similar probability. Participants believed a negative sea rise prediction indicated a value from above the range of expected values more often than a positive prediction. Furthermore, most participants believed that an expert using a negative probability prediction (as recommended by the IPCC) disagreed with other experts, whereas one using an equivalent positive phrase was believed to agree with them.

ABSTRACT. Recruiting staff is one of the most ubiquitous activities carried out in professional contexts. While there is plenty of research on candidate selection (e.g., use of psychometrics), there is very little empirical research on how people make choices between candidates. This paper presents two studies, one looking at short-listing, the other looking at post-interview decision-making. The results of both studies show an extreme vulnerability of recruiters to biases of gender, age, race and disposition, and these biases increase with task complexity. The results also indicate the classic disjunction found in reasoning research between decisions and justifications for them.

ABSTRACT. Exposure to nature decreases negative affect, promotes stress recovery, and improves cognitive performance after mental fatigue. However, higher-level cognitive processes, such as decision making, have been overlooked. Moreover, studies have generally compared natural environments perceived to be restorative, such as woods or rivers, with built environments perceived to be nonrestorative, such as roads with traffic or industrial areas, paying less attention to built environments perceived to be restorative, such as libraries or museums. We examined whether audiovisual exposure to potentially restorative natural and built (vs. potentially nonrestorative built) environments would improve the ability to apply decision-making rules. Fatigued participants completed parallel versions of the Applying Decision Rules task (and measures of attentional control, self-reported fatigue, motivation, and emotional state) before and after being exposed to audiovisual representations of these environments. Decision-making performance improved after exposure to restorative natural environments, remained unchanged after exposure to restorative built environments, and deteriorated after exposure to nonrestorative built environments. Restorative effects on decision-making performance were not mediated by changes in attention control, emotional state, or motivation, but only by changes in perceived fatigue.

ABSTRACT. The novelty of the EduS4EL project lies in addressing challenges associated with climate change through innovative strategies and methodologies rooted in cognitive and communication psychology, including the use of digital tools to enhance awareness of biases, perception of risks, and foster Critical Thinking regarding the polarized arguments prevalent in public discourse. Macchi,L., Caravona,L.: Critical Thinking promotes a better understanding of climate change, as it renders people more adept at uncovering fallacies and rhetorical tricks; it helps formulating correct opinions, based on reliable data and sources, and developing deliberative skills. Martignon,L., Engel,J.: Understanding risks and assessing their consequences is one of the aims to be achieved. Dealing with trustworthy data on environmental phenomena becomes essential requiring the construction of adequate digital tools as open-source platforms, gadgets, plugins, and the development of adequate simulations. Baratgin,J., Jamet,F., Jacquet,B., Bourlier,M.: Another digital tool could be the use of chatbots. We investigate its effectiveness in schools on critical thinking, environmental literacy, eco-anxiety, motivation and decision making towards more sustainable behaviors, using nudges and some elements of gamification. Jiménez,M., Martínez,L.: The literature evidences the priority of training students about recycling to increase their level of awareness to generate individual changes. These changes, related to transformations in their social, educational and family environments, result in an improvement of the conditions of our environment.

ABSTRACT. Dual process theories examine the interplay between different systems or processes of reasoning and how they are resolved to generate a response. Different theories propose different mechanisms of interaction, but they all share one limiting property. Namely, overall reasoning performance cannot exceed the maximum performance of any one process. Within current theories, a person may respond intuitively using system 1, in which case accuracy is determined solely by system 1. Or, if deliberate system 2 thinking overrides system 1 then accuracy is determined solely by system 2. Or, system 1 and 2 may generate the same response. Therefore, in all cases the upper limit on accuracy is either the maximum performance of system 1 or of system 2.

We propose an alternative. Using an ecologically valid decision-making task in which participants decide about personally meaningful everyday decisions, we demonstrate that the best choice occurs through a combination of intuitive and deliberate thought. We propose that each of these systems explains unique variance in the outcome. Combining these processes therefore generates a more accurate response than either one individually. This model suggests a novel interplay of reasoning processes that is incompatible with any of the current dual process theories.

ABSTRACT. The hybrid model of dual process theories of reasoning (De Neys 2018) suggests that system 1 produces intuitions about problems which can sometimes compete against each other. If two antagonist intuitions have similar activation strengths, uncertainty emerges and the more deliberate and cognitively costly system 2 is then activated to attempt finding a logical solution to the problem. This suggests that many tasks in which participants usually fail to give a correct solution due to wrong initial intuitions should be correctly solved by participants who develop a more accurate and intuitive representation of the problem, while still remaining in system 1. We explore this idea in two different experiments, using two different problems (The bat and ball problem and the pigeonhole problem). We believe allowing participants to interact with the objects described in the problem statement should allow a more intuitive understanding of its different components, and thus allow participants to have a more accurate representation of the problem. Such a representation should in turn improve participants' performance when solving isomorphic problems. We also expect participants with such representations to be able to find the correct answer in a shorter amount of time.

References: De Neys, W. (2018). Bias, conflict, and fast logic: Towards a hybrid dual process future? In W. De Neys (Ed.), Dual process theory 2.0 (pp. 47–65). Routledge/Taylor & Francis Group. https://doi.org/10.4324/9781315204550-4

ABSTRACT. Popular dual-process theories of creativity assert that the creative thinking process involves a convergent-thinking evaluative phase relying on deliberation. We tested this dual-process approach in the domain of convergent thinking using the Compound Remote Associates test (CRA). We implemented a two-response paradigm wherein participants provided their initial response under cognitive load and time constraints, followed by a final response for which they could deliberate. Our findings indicate that participants, in the vast majority of cases, do not require deliberation to produce correct answers in the CRA. We used transformer models and semantic network modeling to explore differences between items and participants that could explain these results. We show that items with a smaller semantic search space are better solved intuitively, and that participants with a more efficient and flexible semantic memory structure display higher intuitive performances on the CRA. These results suggest that sound intuiting in a convergent thinking task relies on associative processes unfolding effortlessly within semantic memory, without necessarily requiring executive, controlled cognitive processes. We demonstrate how a simple model of spreading activation over a semantic network can account for these findings by restricting the number of concepts visited during the initial response stage. We discuss how these results challenges conventional dual-process theories of creativity.

ABSTRACT. Dual-process theories propose that two qualitatively distinct kinds of processes contribute to human reasoning. Under this view, reasoning judgements made when under time pressure or with fewer cognitive resources available are typically driven by faster heuristic processes, while judgements made with more time and resources are more likely to be influenced by slower analytic processes. In contrast, single-process accounts propose that the same cognitive mechanisms are applied in each case. To distinguish between the two kinds of theories, we instantiated them as signal detection models and tested a core difference between their predictions, using signed difference analysis. This approach was applied to a series of targeted reasoning experiments, including a base-rate and transitive reasoning task with adults, and a deductive reasoning task with children. We did not find evidence that rejects the simpler single-process models in favour of the dual-process models. This work challenges popular dual-process models of reasoning and their widespread application.

ABSTRACT. Metareasoning is a framework by which we can understand how cognitive monitoring and control processes impact decisions. These monitoring and control processes, in turn, enable individuals to evaluate and modify their decision-making processes to suit each individual decision context. In a typical experimental paradigm, metacognitions are measured explicitly by participant self-report. We argue this type of paradigm may make such experiments unlike real-world cognition. We borrow and adapt a methodology from studies of nonhuman primate metacognition in order to compare a typical condition, where metacognitions are measured explicitly by self-report, to an alternative condition, where metacognitions are observed behaviourally. We propose two candidate effects: 1) That controlled behaviour is more strongly linked to metacognitive monitoring processes when individuals are required to explicate said monitoring process, and 2) That monitoring processes change quantitatively given the requirement to self-report their outputs. We find experimental evidence for only the second of these effects: individuals’ metacognitive monitoring processes differ when an explicit self-judgment is required. This work provides valuable insight into the nature of controlled thought in psychology and decision science. Potential effects of explicit monitoring of one’s cognitions, the mechanisms by which these effects might exist, and implications thereof will be discussed.

ABSTRACT. Overconfidence is a prevalent problem and it is particularly consequential in its relation with scientific knowledge: being unaware of one’s own ignorance can affect behaviours and threaten public policies and health. However, it is not clear how confidence varies with knowledge. In this paper published in Nature Human Behaviour, we examine four large surveys, spanning 30 years in Europe and the United States and propose a new confidence metric. This metric does not rely on self-reporting or peer comparison, operationalizing (over)confidence as the tendency to give incorrect answers rather than ‘don’t know’ responses to questions on scientific facts. We find a nonlinear relationship between knowledge and confidence, with overconfidence (the confidence gap) peaking at intermediate levels of actual scientific knowledge. These high-confidence/intermediate-knowledge groups also display the least positive attitudes towards science. These results differ from current models and, by identifying specific audiences, can help inform science communication strategies.

ABSTRACT. Meta-reasoning research typically focuses on self-monitoring of thinking processes. However, accurately estimating others' knowledge is vital for social and educational interactions. For example, teachers who accurately gauge students' knowledge can adapt lessons accordingly. Previous studies have indicated that people often base their judgments of others' knowledge on their own knowledge, adjusting for differences between themselves and others. Yet the impact of one's confidence in their knowledge remains unclear. Addressing this, two experiments were conducted. In the first experiment, 81 Hebrew-speakers chose the correct meaning of 37 infrequent English words, rated their confidence, and predicted the performance of others (peers) with similar linguistic backgrounds. Results revealed a strong association between self-confidence and predictions of others' success. In the second experiment, with 115 participants, the 'others' were introduced as having different linguistic backgrounds (English-speaking parents), and in this case no correlation was observed. These findings imply that individuals' confidence in their own knowledge guides their prediction of others' knowledge only when they perceive those others to be similar to them. The study underscores the role of self-confidence and perceived similarity in shaping judgments about others' knowledge.

ABSTRACT. Perceived expertise reflects learners' perceptions of lecturers' professionalism and credibility. In digital learning environments, the perceived expertise of video lecturers may serve as a contextual cue for metacognitive monitoring. However, the influence of such cues in meta-reasoning remains largely unexplored. The present study aimed to examine how the perceived expertise of the lecturer influences cognitive and metacognitive processes. Utilizing a 2x2 experimental design, the study manipulated perceptions of lecturers' expertise (high/low) and the availability of an opt-out option (with/without a "don't know" choice) in a problem-solving task. Participants learned the subject of binary numbers through instructional videos, followed by a testing phase that included a set of problems and self-assessment measures. Results demonstrated that groups with high perceived expertise had lower accuracy than those with low perceived expertise, suggesting a possible over-reliance on the lecturer's expertise. Confidence judgments did not reflect this effect, indicating an insensitivity of the monitoring process. This resulted in under-confidence only for groups perceiving low expertise. The opt-out option improved calibration, underscoring its potential in enhancing self-assessment accuracy. Overall, the study serves as initial indication for the role of contextual cues in meta-reasoning, and calls for future research to delve into these effects.

ABSTRACT. Research shows that people tend to think of others as biased and of themselves as unbiased. We start by reviewing evidence in support for a metacognitive model whereby, when people experience a conflict between their intuition and their deliberation, they answer in line with their deliberation but think others would answer in line with their intuition. We then study how to correct people’s predictions of other’s attitudes: We told participants we had asked other students belonging to their university to report their attitudes towards certain protected social groups. They were then shown each social group and were 1) asked to estimate the percentage of other students who expressed a negative attitude towards the group, 2) asked how confident they were, 3) shown the actual percentage of other students’ negative attitudes. In the first trial, participants predicted others would have more negative attitudes than what these other participants reported. Yet, as trials progressed, participants became better at predicting others’ attitudes and decreased their confidence in their predictions. We discuss the feasibility of implementing this correction at scale, along with limitations and future studies in this line of research.

ABSTRACT. Are humans intuitive Bayesians? It depends. People seem to be Bayesians when updating probabilities from experience but not when acquiring probabilities from descriptions (i.e., Bayesian textbook problems). Decades of research on textbook problems have focused on how the format of the statistical information (e.g., the natural frequency effect) affects such reasoning. However, it pays much less attention to the wording of these problems. Mathematical problem-solving literature indicates that wording is critical for performance. Wording effects (the wording varied across the problems and manipulations) can also have far-reaching consequences. These may have confounded between-format comparisons and moderated within-format variability in prior research. Therefore, across seven experiments (N = 4,909), we investigated the impact of the wording of medical screening problems and statistical formats on Bayesian reasoning in a general adult population. Participants generated more Bayesian answers with natural frequencies than with single-event probabilities, but only with the improved wording. The improved wording of the natural frequencies consistently led to more Bayesian answers than the natural frequencies with standard wording. The improved wording effect occurred mainly due to a more efficient description of the statistical information—cueing required mathematical operations, an unambiguous association of numbers with their reference class and verbal simplification. The wording effect extends the current theoretical explanations of Bayesian reasoning and bears methodological and practical implications. Ultimately, even intuitive Bayesians must be good readers when solving Bayesian textbook problems.

ABSTRACT. The gambler’s fallacy is the tendency to expect random processes to switch more often than they actually do—for example, to think that after a string of tails, a heads is more likely. It’s often taken to be evidence for irrationality. It isn’t. Rather, it’s to be expected from a group of Bayesians who begin with causal uncertainty, and then observe unbiased data from an (in fact) statistically independent process. Although they converge toward the truth, they do so in an asymmetric way—ruling out “streaky” hypotheses more quickly than “switchy” ones. As a result, the majority (and the average) exhibit the gambler’s fallacy. If they have limited memory, this tendency persists even with arbitrarily-large amounts of data. Indeed, such Bayesians exhibit a variety of the empirical trends found in studies of the gambler’s fallacy: they expect switches after short streaks but continuations after long ones; these nonlinear expectations vary with their familiarity with the causal system; their predictions depend on the sequence they’ve just seen; they produce sequences that are too switchy; and they exhibit greater rates of gambler’s reasoning when forming binary predictions than when forming probability estimates. In short: what’s been thought to be evidence for irrationality may instead be rational responses to limited data and memory.

ABSTRACT. Deception is a threat in domains such as law, politics, and everyday life. While there is plenty of research on people’s ability to detect deception, much less exists on how people update their beliefs given potentially deceptive testimony. We present a Bayesian model for handling deception and highlight some counterintuitive implications. Imagine buying a used car from a salesman who might be unreliable. His report on the car’s quality is a function both of its true quality and his propensity to misinform you. Under reasonable assumptions we show that if the salesman tells you it’s a good car you should increase your belief that the car is good but also increase your belief that the salesman is unreliable. This holds irrespective of your priors about car quality or salesman reliability. In two experimental studies we show that participants depart from the Bayesian model. They either judge that the salesman’s report gives no information or that it increases both the probability that the car is good and that the salesman is reliable. We explore the robustness of this effect across various scenarios and probability manipulations. We argue that people use simplifying assumptions that are intuitively appealing but can lead to error.

ABSTRACT. The new paradigm in reasoning made a shift from binary logic to probabilities, Bayesian especially as a norm of reference to evaluate human rationality. People seem to be quite rational under this scope, only Dual Process Theory of reasoning has made the distinction between two types of reasoning. One is fast and automatic, it is the intuitive process. The other, the deliberative process, is conscious but also slow and cognitively demanding. Therefore the question remains whether this new norm should apply to both processes. In this experiment, we investigate how cognitive load affects the accordance to the Bayesian rules. We explore conditional thinking in a probability judgement task adapted for intuitive responses.

ABSTRACT. Inequality of opportunity (IOp) leads to misallocation of human capital and can affect economies via its impact on individual economic decision making. This paper studies the impact of IOp on investment, using a laboratory experiment. We randomized IOp, then subjects chose to invest in a risky asset or savings. Our results suggest that IOp impacts investment choices only for people who are penalized by their circumstances and only once they learn the impact of IOp on their relative position in the income distribution. This disadvantaged group more often invests and invests higher shares of their earnings than the control and advantaged groups. The fact that both IOp and knowledge of relative position need to be present, for the impact on investment to materialize, points to the importance of peer effects, and social preferences more broadly, for understanding the effects of IOp on individual decision making.

ABSTRACT. In optimal-stopping problems, people encounter options sequentially with the goal of finding the best one; once it is rejected, it is no longer available. Previous research indicates that people often do not make optimal choices in these tasks. We examined whether additional information about the task's environment enhances choices, aligning people's behaviour closer to the optimal policy. Our study implemented two additional-information conditions: (1) a transparent presentation of the underlying distribution and (2) a provision of the optimal policy. Our results indicated that while choice patterns varied weakly with additional information when providing the optimal policy, it did not significantly enhance participants' performance. This finding suggests that the challenge in following the optimal strategy is not only due to its computational complexity; even with access to the optimal policy, participants often chose suboptimal options. These results align with other studies showing people's reluctance to rely on algorithmic or AI-generated advice.

ABSTRACT. To better understand and ameliorate cognitive biases, we have been developing more scalable methods to (a) trace reasoning processes and (b) improve faulty reasoning. First, to make think-aloud protocols less tedious and time-consuming, I partnered with startups to develop web apps that record and transcribe people’s reasoning process. The resulting apps expedited data collection from months to hours. And an experiment revealed that thinking aloud neither helped nor hindered reasoning (compared to a control group). Second, to test whether social reflection can enhance reasoning more than solitary reflection, we developed a web app to automatically facilitate solitary and social reflection with varying financial incentives. Again, data collection took hours rather than months. Results indicated that solitary written reasoning about the best answer didn’t seem to improve initially faulty answers. However, chatting about the best answer with someone who disagreed was much more likely to improve initially faulty answers. Unexpectedly, the dissenting discussions improved faulty answers more reliably than cash performance bonuses! Together these results suggest that solitary thinking aloud and reflection may not be enough to overcome faulty impulses on simple cognitive tasks and may require exposure to alternative mindsets or reasons.

ABSTRACT. We examine how price discounts drive children’s healthy choices in the understudied context of a developing economy. We partnered with UNICEF to launch three field experiments in Panamá among 2,242 children who received discount coupons for healthy products to redeem at the school’s kiosk. The experiments examined the effect of four pillars of price discount promotions on redemption rates and product purchase: What to discount (product selection), how to discount (message design), whom to target (children’s age), and whether to discount again (repetition). We uncovered four new insights: First, price discounts alone effectively increased demand among children 6-11 years old. Second, product selection based on relative price drives opposing post-promotion effects: price promotions were more effective long-term for the moderately priced products but less for the more expensive ones. Third, discount messages that require children to derive final prices were more effective among older children, whereas messages that directly communicate the final price were more effective among younger children. Fourth, repetition can amplify or undermine discounts’ efficacy depending on message complexity and children’s age. Our research offers concrete guidelines for researchers and practitioners and sheds light on price promotion interventions that most powerfully nudge children of different ages to act.

ABSTRACT. Consensus agreement is often a reliable cue to a claim’s epistemic value. But what happens when all the sources in a consensus echo the same information from a single source? This so-called “illusion of consensus" occurs when people are equally convinced by a dependent consensus (repeated statements from a single source) and an independent consensus (corroborated information from multiple sources). Repeating the same, unreliable source biases judgments and spreads misinformation. We propose that there are two non-exclusive ways that an “illusion of consensus” can arise: 1) people may overestimate the value of a dependent consensus because repeated statements are easier to process OR because they infer that information was shared because it is credible, and 2) people may underestimate the value of an independent consensus if they are insufficiently convinced that primary sources are impartial. Across several experiments we tested strategies for reducing the “illusion of consensus” by drawing attention to bias in a dependent consensus or impartiality in an independent consensus. Efforts to increase confidence in an independent consensus were more successful than efforts to undermine a dependent consensus. The results offer key insights into the social inference processes involved in judging information quality.

ABSTRACT. Humans are resistant to changing their beliefs even in the face of disconfirming evidence. The Bayesian brain theory suggests that we should update our beliefs optimally in light of new evidence, but recent research indicates that belief formation is far from the Bayesian ideal. Individuals can exhibit "stronger-than-rational" updating or be resistant to revising their beliefs. The present study proposes a novel paradigm to explore perceptions and preferences for belief updating patterns in stable and dynamic stochastic environments, using an advice-taking paradigm. In an experiment (N=567) based on a fishing task, we introduce three advisor characters representing formal updating models: Bayesian, Volatile and Rigid. We find that participants exhibit higher trust for the Bayesian advisor than the Rigid advisor, in the stable but not changeable environment conditions. In the changeable environment, participants exhibit higher trust for the Volatile advisor, compared to both the Bayesian and Rigid advisors. The findings also suggest that participants own learning closely mimics the pattern of the Volatile model. This study illustrates that people can differentiate between Bayesian updating, and its "stronger-than" and "weaker-than" variations, and exhibit preferences for these updating patterns, in different environment structures.

ABSTRACT. Although research in the area of belief updating has flourished in the last two decades, most of the research in the area does not treat beliefs as part of a complex and interactive network. In this study, we investigate whether people are likely to employ auxiliary hypotheses to avoid updating their mental models in light of conflicting information, in different domains. In Experiment 1, we replicate an unpublished study by Kahneman and Tversky and introducing two additional domain conditions: social and physical (N=119). We prompt the construction of an initial mental model and introduce a level of uncertainty in its validity with some additional information. We ask participants to make a prediction by raking nine possible scenarios in terms of their likelihood. Replicating the findings of the original study, only 37% of responses across all domains invoked the invalidity of the original report as the explanation for the conflict.In Experiment 2, we introduce a manipulation of the credibility levels associated with both the original and additional information. A preliminary pilot study (N=20) showed that when credibility levels are similar (both low or both high), participants are more inclined to generate auxiliary hypotheses to reconcile conflicting information. This research is ongoing.

ABSTRACT. People tend to prefer explanations that have virtues such as simplicity or breadth. We examined whether their preferences are also influenced by controllable aspects of an explanation. We tested participants’ tendency to explain an outcome (a patient has 2 symptoms) by choosing causes (2 diseases) which are within their control. In Experiment 1 (n=174) we provided antecedent information (e.g., smoking causes disease 1) for each of four causes and varied whether the cause was controllable or not (e.g., disease 1 is caused by smoking vs there is no known cause for disease 2). In Experiment 2 (n=197) we provided consequent information (e.g., disease 1 can be treated by surgery) for each of four causes and varied whether the cause was treatable or not (disease 1 can be treated by surgery vs there is no known treatment for disease 2). Participants chose the most satisfactory explanation for the symptoms, from a set of explanations. In both experiments, participants tended to choose the explanation with controllable aspects (diseases that had controllable first causes or treatments) rather than ones with uncontrollable aspects. The results identify a new controllability bias in explanation preferences.

ABSTRACT. One of the underlying assumptions of research insight problem solving is that a correct solution is a marker of conceptual understanding of the problem; there is a strong theoretical link between forming a “correct” representation of the problem solution and problem success. However, problem success often does not transfer either to the same problem at a different time or to an analogous problem suggesting that conceptual understanding varies in strength. In this study, 109 participants took part in an analogical transfer task based on the Cards problem (Cunningham & MacGregor, 2008). We operationalised conceptual understanding by counting the number of time participants re-checked the instructions. We found a strong relationship between problem success and instruction checks for both the source and the target problem. Notably, those participants who failed to transfer problem success at time one to time two checked the instructions significantly more than those who successfully managed to transfer. Those participants who got the problem wrong at time one and then went on to solve at time two also checked the instructions significantly less than those who continued to fail. The findings suggest that understanding of problem structure is key to successful transfer and provides a novel way to measure this.

ABSTRACT. There are two paradoxes implicit in the insight problem-solving literature: how a state of impasse can be at once necessary to models of insight problem-solving and yet also have appear to have a catastrophic effect on solution rates, and why individuals such as problem-solving and gaming enthusiasts seem to seek out this apparently aversive state. We go some way to solve these paradoxes by presenting a model of impasse as multidimensional based on 4 experiments (total N = 856) drawing on qualitative reports and subsequently confirmed through quantitative analysis. We propose that impasse is both dynamic and unstable (it can be resolved or unresolved) and varies in terms of feelings of cognitive speediness, motivation, and affect. We demonstrate that the feeling of insight can be reliably elicited by experiencing and resolving impasse but also in the absence of impasse, which suggests that there is more than one path to an insight experience. Our findings are robust across a range of problem types. The novel conception of impasse as dynamic and multidimensional has implications for theories of insight problem solving, and also wider implications for understanding how impasse can be resolved across different domains such as education and design.

ABSTRACT. How do we search our memory for answers to factual questions? While much literature explores broader memory processes like free association, the specific mechanisms guiding our internal search for answers are less understood. We asked 27 participants to answer 16 trivia questions, and analyzed their responses using natural language processing. We hypothesized that the answers people generate depend on 1) how similar they are to the question, 2) how similar they are to the previously generated answer, and 3) general prominence or accessibility of that answer. We measured semantic distance using sentence transformers and word embeddings, and also employed GPT-4 to rate semantic distance. To measure prominence, we used Google Trends data for the average search frequency of each answer over the six months preceding the study. We discovered that memory search is dynamic; initial answers are more aligned with the question and more prominent, but as more answers are generated, they diverge from the question and become less prominent. Additionally, we found that previous answers didn't significantly influence subsequent ones. This research offers a unique insight into memory search, shedding light on how our goals and available knowledge direct our retrieval process.

ABSTRACT. We report a new computational model of insight problem-solving, PROGIGI (PROgress and Discovery of Ideas in Generating Insights) implemented in ACT-R. The model operates through the application of two simple heuristics; maximisation of progress, and conceptual discovery through seeking ideas in previous attempts. The model is demonstrated with respect to two problems; The variants of nine-dot problem, and the Cards problem, which is an analogue of Mendeleev's discover of the periodic table of chemical elements. We illustrate tests of the model with empirical studies of variants of both problem and contrast the model against alternatives.