United States

Humans can remember arrays or sequences of stimuli that exceed working memory limits in many domains, from auditory sequences to geometric shapes. This ability has been interpreted as evidence for language-like representations that compress stimuli into compact descriptions. We extend this evidence in the domain of geometry by showing that representations of geometric shapes are not only compressed but also syntactically structured. Experiment 1 shows that different representations can be induced for the same geometric shape, indicating structural representation. Experiment 2 shows that subparts of a shape are easier to recognize when they belong to the same subtree than whey do not, indicating hierarchical organization. Taken together, the results indicate that geometric shapes are encoded in representations that possess internal syntax, just like natural language sentences.

CogSci 2025

Constituency tests in human adults’ language of thought for geometry

psychophysics

language and thought

cognitive architectures

psychology

representation

perception

vision

syntax

### 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.

poster

Misinformation poses a growing challenge in society, particularly as people seem reluctant to revise discredited information. However, emerging research suggests that people’s persistence in believing discredited information is sometimes rational, with individuals applying their own assumptions in a consistent way. In this experimental study, we employ a political vignette with a false accusation to show that even in politically charged contexts, people can correct misinformation and return their beliefs to baseline. Our results demonstrate that participants' belief updating generally aligns with Bayesian predictions, although with a more conservative approach. With regards to source evaluation, participants were more likely to downgrade the reliability of an accuser’s claim when it conflicted with their political views but returned to their initial assessments after the correction. This suggests that while worldview can affect source evaluation at the individual level, this effect does not necessarily translate into a broader erosion of institutional credibility. This study enhances our understanding of how worldview influences belief revision and source evaluation, especially in politically sensitive contexts.

The Role of Worldview Congruence in Misinformation Correction: A Bayesian Approach to Belief Updating

Children’s understanding of privacy develops as they navigate both physical and digital spaces. This study examines how children aged 3- to 13- years-old conceptualize privacy through their drawings, analyzing data from the Privacy Illustrated dataset. We explored two key dimensions: regulation (mechanisms controlling access, such as doors) and content (what children consider private, such as bedrooms or intellectual property). Our findings suggest that as children get older, they are more likely to view privacy as something that can be actively managed using physical barriers and control mechanisms. In contrast, younger children often depicted privacy as simply being alone. Content-related depictions remained relatively stable across ages, though older children included more abstract ideas, such as digital privacy. This study provides a novel framework for examining privacy development, highlighting distinct but interrelated dimensions of privacy.

Drawing Privacy: How Children Conceptualize Regulation and Content Across Development

Recent work in Bayesian, agent-based modelling of scientific communities has employed the Bala-Goyal framework to study the mechanisms involved when industry influence applies the so-called 'Tobacco  Strategy' to undermine collective inquiry. Motivated by limitations of these models, we propose an alternative based on a recently introduced framework for normative argument exchange across networks. We implement representations of two distinct types of industry influence: `Obfuscating' influence directs inquiry to experiments with low expected value of information. `Misleading' influence filters private research and only communicates misleading signals from the world. We explored the impacts of both strategies on the polarization \& mean error of, and flow of information through, social networks of scientists via computer simulations. We conclude that even against highly optimistic background assumptions, and in a less simplified model of inquiry and argumentation, industry influence poses a plausible threat to collective deliberation.

Industry Influencing Collective Scientific Reasoning: A Bayesian, Agent-based Exploration

Why do large-scale field experiments in education often have muted effects? Drawing on system dynamics and self-regulated learning theory, we sought answer this question by simulating the behavior of self-regulated (discrepancy-reducing) learners over time affected by different types of educational interventions. We analyze three types of interventions: changing students’ learning rates (learning strategies), intercepts (prior knowledge or teaching effectiveness), and norms of study (achievement goals). We uncover situations where educational interventions can affect achievement in the short run, but typical cross-sectional analyses do not find a measurable effect in the long run. Results indicate that highly motivated, self-regulated learners may resist external interventions, particularly those targeting learning strategies or prior knowledge. In contrast, interventions show the greatest effect on achievement when students are under time constraints and struggling to achieve their desired performance. Ultimately, self-regulated learners may be the hardest to help, a phenomenon we call the "self-regulated learning paradox."

The self-regulated learning paradox: Or, one reason why educational interventions might fail

What factors of our learning experiences enable us to best acquire complex skills? Recent ideas from artificial intelligence point to two such factors: (1) a balance of real experience with simulated experience acquired during planning itself, and (2) appropriate diversity in training examples. To test whether these factors influence the development of human expertise, we analyzed data from 1,873 chess players on the online platform Lichess, each of whom played hundreds to thousands of games over months to years. We found that both the time spent planning before moves and the diversity of opening positions encountered predict skill improvement over time. These findings suggest that principles shaping the development of expertise in artificial intelligence systems may also apply to human learning.

Learning in online chess increases with more time spent thinking and diversity of experience

Efficiency informs perceptions and expectations of people’s actions from early in life. We examined whether young children aged three and four consider efficiency when evaluating how well agents completed goals. In three experiments, we showed children scenarios where two characters each walked to target objects, and then asked children which character did a better job. In the first experiment, children appeared to consider efficiency. They more often chose a character who took a direct path over one who took an indirect path, but only when the latter character could have taken a shorter path. Two follow-up experiments, though, failed to replicate this pattern and the success in the initial experiment could be explained in terms of the features of the paths (not strictly related to efficiency) used in that experiment. The findings suggest, then, that three- and four-year-olds do not yet use efficiency to normatively evaluate actions. We consider two alternative explanations for this.

Evaluating actions: Do young children prefer actions completed efficiently over those completed inefficiently?

How can an AI teammate implicitly coordinate with a human? We address this question by integrating Instance-Based Learning (IBL), a cognitive theory of learning and decision making, with the level-k Theory of Mind framework. We hypothesize that coordination emerges when partners adopt complementary k-levels and when the higher k-level agent has an accurate model of their partner's cognitive processes. To test this hypothesis, we introduce a simultaneous-choice, multi-attribute task, where outcomes depend on interactions between choice features and agent decisions. Simulations of pairs of IBL-based agents at different k-levels support the hypothesis that complementary k-levels enhance coordination. However, empirical results from an experiment reveal no advantage of [human, IBL-L2] pairs over [human, IBL-L1] pairs, even when participants are restricted to operate as L1 agents. Post-hoc simulations show that model fitting recovers some advantage for [human, IBL-L2] teams by enabling the IBL-L2 agent to more accurately predict their human partner's actions.

Facilitating Human-AI Coordination through Computational Theory of Mind

This study was the first to examine whether the volume of one’s voice serves as an embodied cue for assessing information credibility. Eighty U.S. undergraduate students were randomly assigned to one of three conditions: loud, soft, or control. They read aloud in their assigned loudness condition while rating the truthfulness of trivia statements, followed by silently rating additional statements. The results revealed no significant effect of voice loudness on truthfulness ratings. When examining confidence levels reflected in the ratings, an interaction effect between reading status and loudness condition emerged. Participants who controlled their volume (either loud or soft) rated statements with higher confidence compared to when rating statements silently. These findings suggest that speakers do not associate their own voice loudness with the truthfulness of information.

Amplifying Truth? Vocal Volume and Speakers’ Self-Perceived Truthfulness

Prediction is essential in language processing and is often studied as an isolated task. However, in real life, listening often occurs alongside other cognitively demanding activities. We investigated how cognitive demand affects prediction in older adults and whether hearing impairment impacts this process. Eighty-four older adults with hearing impairment (HI) or normal hearing (NH), matched in working memory scores, participated in a visual world paradigm study with semantically predictable and unpredictable sentences. The experiment featured two conditions: listening without a memory task and listening while concurrently retaining a three-item visuo-spatial memory load to simulate competing cognitive demands. Divergence point analyses showed that memory load delayed prediction timing in NH participants, and an interaction revealed differences occurring with hearing loss. As prediction timing was impacted both by load and hearing loss, our work suggests that prediction requires executive resources.

One Thing at a Time: Investigating the Impact of Increased Cognitive Demand on Semantic Prediction in Older Adults with and without Hearing Impairment

Prediction is central in human language processing, as the brain continuously predicts upcoming words using prior knowledge and context. Surprisal theory quantifies predictability using word surprisal. While previous studies link neural activity to surprisal during passive listening or reading, we investigate how surprisal is tracked in dynamic face-to-face conversations. Two key challenges arise: estimating surprisal as well as identifying predictions in EEG data in natural conversation. We address the first challenge by adapting a pre-trained large language model to a dataset of spontaneous conversation capturing features like hesitations and repetitions. We then relate the surprisal estimated by the adapted model to EEG data using temporal response functions. Our experimental results show neural tracking of surprisal at different time lags after word onset, supporting the surprisal theory in face-to-face conversation. To the best of our knowledge, we are the first to address the application of surprisal theory in such interactive settings.

Downloads

Next from CogSci 2025

The Role of Worldview Congruence in Misinformation Correction: A Bayesian Approach to Belief Updating

Similar lecture

Hierarchical Abstraction Enables Human-Like 3D Object Recognition in Deep Learning Models

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES