United States

What kind of thing is a theory of mind? How can we formalize the various theories of mind we study as cognitive scientists? In this paper, we argue that it is valuable to think of a theory of mind as a kind of programming language: one that is specialized for setting up and reasoning about problems involving other minds. Drawing on ideas from the theory and history of programming languages, we show how this perspective can help us formalize concepts in a theory of mind, precisely articulate differences between multiple theories of mind, and reason about how we develop our theories-of-mind over time. (The full version of this paper is available at: https://kach.github.io/memo/tomalot)

CogSci 2025

Theories of Mind as Languages of Thought for Thought about Thought

social cognition

theory of mind

computational modeling

computer science

bayesian modeling

The purpose of the International Conference on Agents and Artificial Intelligence is to bring together researchers, engineers and practitioners interested in the theory and applications in the areas of Agents and Artificial Intelligence. Two simultaneous related tracks will be held, covering both applications and current research work. One track focuses on Agents, Multi-Agent Systems and Software Platforms, Distributed Problem Solving and Distributed AI in general. The other track focuses mainly on Artificial Intelligence, Knowledge Representation, Planning, Learning, Scheduling, Perception Reactive AI Systems, and Evolutionary Computing and other topics related to Intelligent Systems and Computational Intelligence.

ICAART 2020

The purpose of the International Conference on Agents and Artificial Intelligence is to bring together researchers, engineers and practitioners interested in the theory and applications in the areas of Agents and Artificial Intelligence.

poster

### Welcome to CogSci Conference 2025!

The 47th Annual Meeting of the Cognitive Science Society was a hybrid meeting held in San Francisco. 

<div style="position:relative;padding-top:0;width:900px;height:500px;"><iframe style="position:absolute;border:none;width:100%;height:100%;left:0;top:0;" src="https://online.fliphtml5.com/ebtyf/amvr/"  seamless="seamless" scrolling="no" frameborder="0" allowtransparency="true" allowfullscreen="true" ></iframe></div>

#### About

The Cognitive Science Society brings together researchers from around the world who hold a common goal: understanding the nature of the human mind. The mission of the Society is to promote Cognitive Science as a discipline, and to foster scientific interchange among researchers in various areas of study, including Artificial Intelligence, Linguistics, Anthropology, Psychology, Neuroscience, Philosophy, and Education.

The Society is a non-profit professional organization and its activities include sponsoring an annual conference and publishing the journals Cognitive Science and TopiCS.

#### Our History 

* **Society Creation**<br>
The Society was incorporated as a 501(c)(3) non-profit professional organization in Massachusetts in 1979. The organizing committee included Roger Schank, Allan Collins, Donald Norman, and a number of other scholars from psychology, linguistics, computer science, and philosophy. 
<br><br>
* **Conference Creation**<br>
The first conference on cognitive science was held at La Jolla, California in August, 1979, and has occurred annually since then. The proceedings of each conference are published, and those from most years are available through Lawrence Erlbaum Associates, Inc. The annual proceedings of the Cognitive Science Conference represent a major source of information on new work and new ideas in the scientific study of thinking. In 1990, the Society, with help from an anonymous donor, established the David Marr Prize for the best student paper at each annual meeting.
<br><br>
* **Journal Creation**<br>
The Journal, Cognitive Science, began publication in 1976, and is now published by Wiley-Blackwell. The Executive Editor is currently Richard P. Cooper of Birkbeck, University of London, and there are 18 Associate Editors and a 30-member editorial board. It serves as the premier outlet for research reports that intersect two or more disciplines. Copyrights for articles published in the journal are held by the Society. The Governing Board of the Cognitive Science Society voted in late 2006 to found a new journal, Topics in Cognitive Science (topiCS). The Editor in Chief is Wayne Gray, Cognitive Science Department, Rensselaer Polytechnic Institute. The journal seeks to fill a niche not occupied by Cognitive Science Journal or other cognitive science journals. Membership in the Society includes a subscription to Cognitive Science and TopiCS. Copyrights for articles published in the journal are held by the Society.
<br><br>

#### Code of Conduct

By attending the CogSci 2025 Conference, you are required to adhere to the society’s **[Code of Conduct](https://drive.google.com/file/d/1ChPuihLy6jE_BWqfO7J2KKgX35JW2zsM/view?usp=sharing)**.
<br><br>


You need to log in with the email address you registered with. 

Login credentials were sent to you from Underline -  subject line "Welcome to the CogSci 2025 Conference". Please be sure to check your spam/promotional inbox  if you do not see an email confirmation right away.





Please log in to join this event.

To access the site, please register [**here**](https://cognitivesciencesociety.org/registration/).

If you are registered and feel like you are seeing this message by mistake, please make sure you are logged in with the same email that you registered with. 

Please register!

The 47th Annual Meeting of the Cognitive Science Society presents the latest research across cognitive science and highlights the theme of Cognition in Context.

This study compares classification and observation training in categorization tasks involving multiple rules, including an optimal XOR rule and suboptimal uni-dimensional rules. Participants (N = 192) were assigned to either condition, with classification involving active categorization and feedback, while observation involved studying the pair of category label and item together. Results showed that classification participants outperformed observation participants in accuracy and exhibited greater consistency in strategy use. Bayesian modeling revealed no significant difference in rule adoption between conditions, but classification led to fewer strategy switches and lower error rates. These findings suggest that classification training enhances performance by fostering stronger commitment to adopted strategies. The study highlights the importance of strategy commitment in categorization and questions the reliance on overall accuracy alone as a performance metric.

Training Methods in Categorization: A Comparison of Classification and Observation on Rule Adoption and Rule Consistency

Humans predict upcoming language input from context, which
depends on prior language experience. This suggests that older
adults’ predictions may differ from those of young adults, due
to longer language exposure. Here we use sentence completion
data from two age cohorts (YA = 18-35 y.o.; OA = 50-80
y.o.) and language models fine-tuned to particular decades of
a diachronic corpus of American English to examine the relationship 
between changes in language statistics and differences in 
linguistic prediction across different age groups. We
observed greater consistency in contextual probabilities within
age groups compared to across age groups, indicating that
YA and OA make subtly different predictions given identical
context. Next-word prediction performance for the fine-tuned
models decreased as the temporal distance between the fine-
tuning and testing decade increased, indicating that language
usage statistics changed over the span of a few decades. Further, 
GPT-2 surprisal values are more predictive of YA than OA
contextual probabilities, suggesting that the language statistics,
as captured by a model trained largely on internet text, aligns
more with YA’s internal model than OA’s. However, both age
groups’ data are better fit by models fine-tuned on more recent
corpus decades.

Language experience and prediction across the lifespan: evidence from diachronic fine-tuning of language models

Failure is an unavoidable part of learning, and how to manage it influences children’s future learning and success (Eskreis-Winkler and Fishbach, 2019). Whereas children can experience both negative and positive emotions following failure (Nelson et al., 2018), it remains unclear how children’s achievement emotions change following both failure and success, and whether emotional change aligns with future learning choices. A total of 107 preschool children attempted three unsolvable puzzles without help, then successfully completed one solvable puzzle with guidance as needed. ANOVA results showed a significant negative linear trend in children’s affective responses, with no rebound after the success experience. When selecting a future learning task, children exhibited a strong preference for the prior successful rather than any of the incomplete puzzles; yet, logistic regression indicated that this decision was not predicted by changes in emotion. Findings suggest that children showed an adaptive emotional response following both failure and success.

Learning from Failure and Success: Children’s Achievement Emotions and Learning Choices 

We investigated human learning of probabilistic rules in experiments using a 2-by-2 feature space. Despite the seemingly minimal complexity, participants struggled to learn nonlinear XOR rules (where outcomes depend on cue matchings) but rapidly mastered linear rules. This difficulty persisted even when explicit probes revealed the possible rules, indicating constraints in hypothesis testing. To explain these behavioral patterns, we propose a hypothesis diffusion model where learning arises from evidence-driven transitions between hypotheses in a sparsely connected network. The model outperformed reinforcement learning alternatives and generalized across different rules. To further understand the origin of the learning difficulty, we trained low-rank recurrent neural networks and found that networks with limited capacity (rank 3) failed to learn XOR rules when trained in biased environments, mirroring human performance. In conclusion, human rule learning may rely on structured hypothesis exploration, with learning biases potentially emerging from adaptations to environmental demands under computational constraints.

Bounded hypothesis testing underlying human learning of probabilistic rules

Our knowledge of a topic such as mathematics is reliant upon the hierarchies of chunks we build in our memories.  The time course of knowledge-based tasks, such as the transcription of algebraic formulas, provides rich signals that reflect the structure of the chunk hierarchies being processed.  By building Gaussian Mixture Models, this paper provides evidence that decomposing the overall dis-tribution of pauses between actions in a sequential task can give meaningful characterizations of the structure of the chunk hierarchy.  We also examine whether individual competence in mathematics can be measured using a metric derived from the models.

Evaluating the Structure of Chunk Hierarchies in a Naturalistic Educational Task Using Gaussian Mixed Models

Determining what caused an event is common in everyday life, yet little is known about what aspects of real-world events affect causal attribution. Causes may unfold at multiple timescales, with gradual events (e.g., steady weight loss) and abrupt ones (e.g., acute illness) contributing to an outcome. We investigated causal attribution in real-world contexts (e.g., finance, health) where both types of events contribute to a positive or negative outcome. In Study 1, participants gave higher causal ratings to abrupt causes in negative scenarios about the environment or finance. Conversely, we found higher causal ratings for gradual causes of physical health regardless of outcome valence, and no significant differences in mental health scenarios. Further, participants' counterfactual responses were mostly consistent with their causal attributions. Study 2 suggests that the preference for abrupt causes may be explained by their temporal proximity to the outcome. We discuss explanations for these findings and their implications.

Causal and Counterfactual Reasoning about Gradual and Abrupt Events

The current study utilizes the Ganzflicker paradigm—a flickering stimulus that induces visual hallucinations—to provide insight into the internally-generated visual experiences that correlate with individual differences in visual imagery. Here, we analyzed rich narrative descriptions of Ganzflicker hallucinations from 4,365 participants using natural language processing, sensorimotor norms, and AI visualizations. We find that overall perceptual richness and visual detail in descriptions increase with imagery vividness. Examining the specific content of these descriptions reveals that vivid imagers report more face and hand-related content than those with weaker imagery. Exploratory AI-generated visualizations of these descriptions provide additional insights, as those with weak imagery report patterns of simple visual features, like colors and geometric forms, while strong imagers’ hallucinations are filled with complex real-world stimuli. These findings suggest imagery differences may lie not in early visual processing but in the integration of basic visual features into complex object- and scene-level representations.

Visual Imagery Vividness Predicts the Complexity of Induced Hallucinations 

Number concepts are often thought to originate from counting and the successor function, or from a refinement of the approximate number sense. Here we argue for a third origin: a shared language-of-thought for geometry and arithmetic, with primitives of repetition, concatenation, and recursive embedding. Applied to sets, starting from 1, those primitives engender concepts of exact integers through recursive applications of additions and multiplications. Links between geometry and arithmetic also explain the emergence of higher-level notions (squares, primes, etc.). Under our hypothesis, understanding a number means possessing one or several mental expressions for it, and their minimal description length determines how easily they can be mentally manipulated. Several historical, developmental, linguistic, and brain-imaging phenomena provide preliminary support for our proposal.

Origins of numbers: A shared language-of-thought for arithmetic and geometry?

We report a large study examining the effects of tDCS on the FIE. Subjects randomly assigned to one of the four tDCS groups and engaged with an old/new recognition task involving upright and inverted faces. 1) Sham tDCS during the study phase and recognition task; 2) Anodal tDCS during the study phase followed by sham tDCS during the recognition task; 3) Anodal tDCS during the study phase followed by cathodal tDCS during the recognition task; 4) Cathodal tDCS during the study phase followed by sham tDCS during the recognition task. Group 2 confirmed that anodal tDCS reduces the FIE vs. sham (Group 1) by disrupting performance for upright faces. Group 3 showed that cathodal tDCS applied after anodal, increased the FIE vs. Group 2, bringing it back to sham level, by enhancing upright faces.  Group 4 revealed that cathodal tDCS applied after sham has no effect on the baseline FIE

Applications of transcranial Direct Current Stimulation (tDCS) for modulating the Face Inversion Effect (FIE): Reducing and Enhancing Recognition

Why do we give the explanations we do? Recent work has suggested that we should think of explanation as a kind of cooperative social interaction, between a why-question-asker and an explainer. Here, we apply this perspective to consider the role emotion plays in this social interaction. We develop a computational framework for modeling explainers who consider the emotional impact an explanation might have on a listener. We test our framework by modeling human intuitions about how a doctor should explain to a patient why they have a disease, depending on the patient's propensity for regret. Our model predicts human intuitions well, better than ablations suggestive that people do indeed reason about emotion when giving explanations. See https://sites.google.com/view/empathy-in-explanation for further details and pre-print.

Downloads

Next from CogSci 2025

Training Methods in Categorization: A Comparison of Classification and Observation on Rule Adoption and Rule Consistency

Similar lecture

Testing and Training Theory of Mind for Hybrid Human-agent Environments

.css-70qvj9{display:-webkit-box;display:-webkit-flex;display:-ms-flexbox;display:flex;-webkit-align-items:center;-webkit-box-align:center;-ms-flex-align:center;align-items:center;}Downloads

Next from CogSci 2025

Training Methods in Categorization: A Comparison of Classification and Observation on Rule Adoption and Rule Consistency

Similar lecture

Testing and Training Theory of Mind for Hybrid Human-agent Environments

Downloads