United States

What and how do people learn about visual appearance from language? We test the hypothesis that in the absence of sensory evidence, people born blind use abstract causal knowledge to infer object appearance. Congenitally blind (n=19) and sighted adults (n=59) reported how many colors two types of artifacts were likely to have: artifacts for which having many colors is intended to facilitate function (n=30, e.g., fairytale book, fruit candies), and artifacts for which colorfulness is irrelevant or distracting (n=30, e.g., instructional manual, painkillers). The number of colors estimated per object was highly correlated across groups. Blind and sighted people assigned more colors to artifacts for which colorfulness facilitates function and appealed to makers’ intentions in open-ended explanations. A text-only version of GPT-4 generated similar but non-identical colorfulness estimates compared to humans. Our findings suggest that people infer the appearance of unseen objects using causal ‘intuitive theories’ informed by linguistic evidence.

CogSci 2025

Learning visual appearance from language is mediated by causal intuitive theories

cognitive development

language and thought

concepts and categories

intelligent agents

causal reasoning

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.

We explore how people “bootstrap”, or reuse chunked action sequences, to tackle complex problems, in a novel puzzle task. In this task, participants perform sequences of actions to recreate target shapes. In our experimental condition, participants are trained on problems whose best solutions share a distinct abstract action sequence, or schema. Meanwhile, a control group trained on tasks of commensurate difficulty whose solutions did not conform to this pattern. We find experimental-condition participants outperform controls in a set of more difficult test puzzles whose solutions are compositional generalizations of experimental group’s training tasks. Notably, the experimental group outperformed controls even in “far transfer” tasks that lacked surface similarity to training in both their target shape and solution sequence. Our results provide a compelling demonstration of the human ability to cache and reuse abstract patterns, offering new insights into how humans approach complex problems that, naively, seem to demand a prohibitive amounts of planning or trial and error.

Bootstrapping in Geometric Puzzle Solving

During conversation, individual speakers become part of a larger conversational system. This is observable in interpersonal coordination, variously described as alignment, synchrony, convergence, and complexity matching. While this coordination is ubiquitous, the degree of coordination varies greatly between conversations, interlocutors, and group-level goals. There is growing interest in understanding the constraints that operate at the level of the conversational system. We explore how social distance, the status and demographic differences of conversation partners, constrain linguistic alignment. To do this, we leverage the CANDOR dataset (Reece et al., 2022), which contains 1656 approximately half-hour ZOOM conversations between strangers and a large battery of follow-up questions regarding their conversation, personality, and demographic information including perceptions of status. In this exploratory study, we use recurrence analysis to examine how various coordinative behaviors, particularly lexical and syntactic alignment, are influenced by the social distance between conversation partners. Experimental follow-ups are discussed.

Social position as a constraint on linguistic alignment

The human brain supports complex behaviors through diverse functional connectivity patterns. We propose Pareto optimality as a novel framework to understand this functional organization. According to Pareto theory, systems optimizing multiple competing goals do so by balancing trade-offs along a low-dimensional "Pareto front" defined by archetypes that each optimize a single goal. Applying Pareto analysis to resting-state fMRI data (HCP, N=1200), we found that individual connectomes lie on a low-dimensional triangle. The three archetypes represent core computational goals: minimizing energetic cost, supporting cognitive control and goal-directed behavior, and enabling internal processing and memory. These goals are reflected in connectivity patterns, network topology, information flow, behavioral and clinical associations. The framework generalizes beyond rest to task-based brain states, and a simple neural model illustrates the trade-offs’ computational basis. Pareto optimality offers a principled approach to decompose brain function into core computations across conditions, populations, and stages of life.

Pareto optimality reveals the core computations of the human brain

Empathy is an emotion that plays a key role in emotional understanding and perspective-taking, and has been identified as a strong motivator for prosocial behavior. We explore people's intuitive theory of empathy, focusing more specifically on the role that the concept of empathy plays in people's causal model of prosocial behavior. We suggest that people implicitly think of empathy as indexing the weight that the actor puts on the welfare of the recipient when deciding whether to help. We test this proposal by asking participants (N=150) to read a series of vignettes in which an actor has the opportunity to help a recipient in need. We find that participants have a robust expectation that actors who feel empathy for the recipient are more likely to help. Furthermore, participants seem to expect that actors who feel empathy are more sensitive to the potential benefits of an action when deciding whether to help. We also test if people can `invert' this intuitive theory to make inferences about an actor's empathy, given their observable behavior. We find only weak evidence that they can do so, although this might be due to limitations in our experimental design. Overall, our work is a first step toward elucidating the computational principles underlying laypeople's conception of empathy.

Exploring the Intuitive Theory of Empathy 

Prototypicality and frequency have been assumed to play a critical role in children’s sentence comprehension. However, Mandarin-speaking five-year-olds’ comprehension of the SVO structure consisting of subject verb object, ba structure consisting of subject ba object verb, and bei structure consisting of subject bei object  verb, all of which were plausible but contrasted with animacy, namely animate-animate (AA), animate-inanimate (AI), inanimate-animate (IA), and inanimate-inanimate (II) indicateed that the prototypical AI combination did not surpass all the remaining patterns and the ba and bei structures with low frequencies did not necessarily result in worse performance. Animate noun phrases (NPs) did not always trump inanimate NPs to be agent/doer. Although five-year-olds’ performance with the IA bei structure followed eADM’s and good-enough representation’s predictions, the ba structure’s better comprehension in the AI and II conditions than any other structure suggests that arrangement of argument order may not be as prominent as those theories claimed.

Prototypicality and frequency do not necessarily affect children’s sentence comprehension 

Children's language processing differs from adults' in idiosyncratic ways. Within adult data, various incremental processing phenomena have been shown to be predicted by neural language models (LMs) using surprisal as the linking hypothesis, where processing effort is determined by a word’s log inverse conditional probability. Since LMs seem without strong inductive bias for natural-language-specific structures, with their word predictions determined by training on naturalistic data, these results potentially support exposure-based theories. However, it remains unclear how well LMs explain the developmental trajectory of human language processing. Here we evaluate LMs with developmentally-realistic training data—how well they predict six established child language processing phenomena, including cases where child and adult patterns differ. Our LMs correctly predict four of the six, but fail in cases involving thematic-role and pragmatic knowledge. Our results highlight the limitations of language-exposure-based theories and call for further empirical research on human language processing patterns throughout development.

Surprisal and developmental sentence processing: exploring the role of language exposure through neural language models

Proportional reasoning is critical for successful functioning across domains and development. However, proportional information is also complex, resulting in behavioral variation across contexts and tasks. In the current study, we systematically compare adults’ proportion judgements on a proportion magnitude comparison task and an equivalent proportion matching task with both dot arrays and continuous rectangles. We find that the match-to-sample task is more difficult than the magnitude comparison task and dot arrays are more difficult than the rectangles. Interactions between task and format, as well as specific patterns of errors, provide additional insight into possible explanations for these patterns. Overall, findings provide theoretical insight into the cognitive processes involved in solving proportional tasks and methodological insight into how to best design and interpret performance on both comparison and match-to-sample proportion tasks.

 Variation in Adults’ Judgements about Relative Proportional Magnitude and Proportional Equivalence

A central question in cognitive science is the degree to which human and animal brains have adapted to and internalized the physical laws that govern the motion of objects. In this project, we propose a new method to estimate aspects of our intuitive sense of physical laws. Rather than assuming that humans internalize the form of Newtonian physics as found on Earth, we instead designed a procedure which allowed us to estimate which forms of physical laws feel most natural and intuitive to human participants.  Our approach combines Markov chain Monte Carlo with People (MCMCP) and a custom parameterized physics engine.  Each proposal of the MCMCP chain instantiated a world with new physical parameters and participants judged which of two scenes seemed more natural. Preliminary results show that this approach can quantify the precision of people's estimate of the direction and strength of gravity.

Estimating Intuitive Physical Parameters using Markov Chain Monte Carlo with People

No matter how brilliant, one person cannot achieve major technological innovations alone. Human progress relies upon our ability to think together, building beyond an existing foundation of cumulative cultural knowledge (Heinrich & Muthukrishna, 2024). From five-years-old, children show cooperative capacities fundamental to this collective success (Warneken et al., 2014; Fletcher et al., 2012). Yet, little is known about children’s capacity to pool mental resources with cooperative partners – if they can think together as interconnected nodes to surpass individual computational limits (Velez et al., 2022). Prior developmental research also does not fully address cross-cultural diversity in children’s cooperative strategy (Rogoff, 2014). Here, we investigate how pairs of children (N = 96 dyads) cooperate on a memory task across two cultural contexts – Nanyuki, Kenya and Beijing, China. We find that children flexibly employ different strategies based on the level of cognitive demand, pointing to an early capacity for strategic cognitive collaboration.

Children’s division of cognitive labor: Evidence from Kenya and China

How do children learn the appropriate scope of linguistic generalizations?  One proposal is that prediction error and cue competition enable them to implicitly reduce their uncertainty about the various cues to linguistic patterns. Previous work has employed artificial language studies to test the predictions of error-driven models against the performance of (adult) human participants (Ramscar et al., 2010). A critical prediction of these models - that linear relations between linguistic and environmental cues can critically affect generalization - has received much empirical support.  For example, Vujovic et al. (2021) found that suffixing languages supported the learning of discriminating cues, and overgeneralization avoidance, better than equivalent prefixing languages. The current study addresses a limitation of previous studies: the use of unnatural flat distributions, which contrast to the skewed distributions ubiquitous in natural language. Although some of our results are consistent with model predictions, there were divergences. Possible reasons for these are discussed.

Downloads

Next from CogSci 2025

Bootstrapping in Geometric Puzzle Solving

.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

Bootstrapping in Geometric Puzzle Solving

Downloads