United States

Statistical learning (SL), or the ability to unintentionally extract patterns underlying sensory information, allows the human mind to acquire various regularities in written languages. However, we are unclear about its potential in learning multiple probabilistic regularities in a complex language, such as the sub-lexical mappings between orthography (form), phonology (sound), and semantics (meaning) in Chinese characters. We tested (implicit) SL and its explicit form in acquiring the Chinese sub-lexical mappings as a randomized controlled trial. Ninety-five 1st-4th graders with or at risk for dyslexia were randomly assigned to an implicit-SL group that was exposed to a set of characters with the sub-lexical form-sound-meaning mappings, an explicit-SL group that was exposed to the same set of characters with explicit instruction on the form-sound mappings, or a no-SL control group. Only the explicit-SL group showed abstraction of form-sound mappings, while only the implicit-SL group showed optimized reading processes across phonology and semantics.

CogSci 2025

Abstraction and Optimization in Statistical Learning: A Randomized Controlled Trial of Implicit and Explicit Reading Intervention for Students with Dyslexia

clinical methods

statistical learning

psychology

reading

education

### 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

The Abstraction and Reasoning Corpus (ARC) is a visual program synthesis benchmark designed to test out-of-distribution generalization in machines. Although recent advancements have led to human-level performance on ARC, it is still unclear how people solve ARC tasks. Other work has demonstrated that people often make errors when reasoning about these problems, and sometimes fail to infer the true underlying program. In this work, we explore the hypothesis that the cognitive mechanisms supporting reasoning are more approximate, graded and resource-limited compared to those suggested by purely discrete, program-induction models. Taking inspiration from previous work on program sketches and partial programs, we model human reasoning in the ARC domain using a neurosymbolic, meta-learned model that interleaves symbolic operations with approximate, statistical pattern completion. We then evaluate the model against human errors from H-ARC, a recent, comprehensive behavioral dataset on the training and evaluation sets of ARC.

A Neurosymbolic Model of Human Reasoning on the Abstraction and Reasoning Corpus

When learning a task by observing another person performing it, an individual can either focus on imitating the other’s behavior (policy imitation), or attempt to infer the other’s goals and beliefs and adjust their own behavior accordingly (goal emulation). Imitation is considered to be computationally cheap but less accurate, while emulation is considered to be computationally costly but more accurate. Drawing upon research on computational resource rationality, we ask whether individuals incorporate cost-benefit considerations when choosing whether to imitate actions or emulate goals. To answer this question, we used an observational-learning extension of a two-step bandit task, and manipulated the reward at stake. Participants’ behavior was best fit by a dual-process model of goal emulation and one-step imitation, consistent with findings from previous research. However, contrary to our hypothesis and inconsistent with cost-benefit arbitration, we found no evidence that rewards at stake influenced participants’ social learning strategies.

No Evidence for Cost-Benefit Arbitration Between Social Learning Strategies

Chinese has many distinct characteristics that differentiate it from alphabetic languages, making it necessary to develop specific theories and models for its study. Previous research on Chinese lacks systematic computational models for lexical and semantic processing. To address this gap, we built a computational model that simulates the processing of Chinese words presented in isolation. The model can process both single-character and multi-character words. Moreover, it can simulate orthographic, phonological, and semantic processing of words, as well as their interactions. These research findings will help clarify the cognitive mechanisms of Chinese reading and deepen our understanding of human language processing. The established model can guide experimental research and has significant theoretical significance.

A Computational Model of Chinese Word Processing

As AI tools become integral to daily life, evaluating their impact on human cognition and learning is essential. This study examines how AI-assisted writing (AIW) tools influence language development over six months. Participants were randomly assigned to either a control group, using basic auto-correction (e.g., Grammarly), or an experimental group, using dialog-based large language model (LLM) tools (e.g., ChatGPT). Each month, all participants will write an English essay and reported on tool usage frequency and strategies. Preliminary findings after 1 month suggest diverse usage behaviors with AIW tools which may lead to varied outcomes in linguistic performance. These results provide insights into AI’s role in fostering language growth and inform strategies for effective AI-enhanced writing practices. In July, we will be able to report 6 months of data.

Does using LLMs in daily life help or hinder learning a second language?

Evaluating the cognitive abilities of Vision-Language Models (VLMs) is challenging due to their reliance on spurious correlations. To distinguish shortcut-taking from genuine reasoning, we introduce Concept Hacking, a paradigm manipulating concept-relevant information to flip the ground-truth but preserving concept-irrelevant confounds. For instance, in a perceptual constancy test, models must recognize that a uniformly wide bridge does not narrow in the distance; the manipulated condition using concept hacking altered the bridge to actually taper. We assessed 209 models across 45 experiment pairs spanning nine low-level cognitive abilities, encompassing all five core knowledge domains. Comparing performance on manipulated versus standard conditions revealed that models fell into shortcut-reliant or illusory understanding types, with none approaching human-level performance. Models of varying sizes appear in each category, indicating that scaling neither imparts core knowledge nor reduces shortcut reliance. These findings highlight fundamental limitations in current VLMs, reinforcing concerns about their ability to achieve genuine understanding.

Evaluating Vision Language Models Through Concept Hacking

Intrusive memories are disruptive to daily functioning and detrimental to well-being; unfortunately, the presence of these memories is a defining characteristic of post-traumatic stress disorder (PTSD). Although increasingly understood as a memory disorder, current trauma management and recovery strategies for PTSD do not often take principal memory theories into account. Many common practices, such as delayed processing after trauma exposure and spaced therapy sessions, might inadvertently strengthen the retention of intrusive memories in the long-term. In this paper, model simulations show that altering the timing of different presentations of emotional stimuli might affect subsequent intrusions. Experimentally, we demonstrated through a two-day within-subject image presentation task that when emotional images are presented in spaced intervals (as opposed to consecutive, “massed” presentation), the perceived frequency of intrusions for the mass-presented emotional images during the 24 hours after first exposure were significantly lower than spaced images. Our study presents a novel strategy that can potentially mitigate the frequency of intrusive post-traumatic memories, highlighting the advantages of translational applications of computational cognitive models to mental health.

Reducing Traumatic Memory Intrusions by Timing Their Re-Encoding: An Application of Computational Modeling to Mental Health

The role of mental simulation in human physical reasoning is widely acknowledged, but whether it is employed across scenarios with varying simulation costs and where its boundary lies remains unclear. Using a pouring-marble task, our human study revealed two distinct error patterns when predicting pouring angles, differentiated by simulation time. While mental simulation accurately captured human judgments in simpler scenarios, a linear heuristic model better matched human predictions when simulation time exceeded a certain boundary. Motivated by these observations, we propose a dual-process framework, Simulation-Heuristics Model (SHM), where intuitive physics employs simulation for short-time simulation but switches to heuristics when simulation becomes costly. By integrating computational methods previously viewed as separate into a unified model, SHM quantitatively captures their switching mechanism. The SHM aligns more precisely with human behavior and demonstrates consistent predictive performance across diverse scenarios, advancing our understanding of the adaptive nature of intuitive physical reasoning.

A simulation-heuristics dual-process model for intuitive physics

Problem-solving often requires both acting on and gathering information from the environment. Prior work in psychology proposes that people are intrinsically motivated to seek information (Ryan and Deci, 2000; Ruggeri et al., 2021). However, in realistic settings most available information has no utility, so optimal performance requires estimating its value. We introduce a model that applies the Bayesian framework of Lidayan et al. (2024), which formalizes the value of information as the resulting increase in expected rewards. Our model calculates the utility of information-gathering actions and treats humans as noisy utility-maximizers. We design a novel task in which information sources vary in their likelihood of enabling downstream rewards. Preliminary results suggest our model predicts human behavior more accurately than intrinsic motivation models, suggesting that humans learn to estimate the value of information from experience, and use it to make better decisions.

Bayes-Adaptive Information Gathering 

For some learning problems, information is readily available. For others, substantial time and effort is needed to acquire information. The present work tests whether this variation in “information accessibility” affects curiosity. In two experiments, we prompted adult participants to rate their curiosity about the answers to trivia questions. For each trivia question, participants were informed that information accessibility would be high—they would receive the answer with minimal time and effort—or low—they would receive the answer with substantial time and effort. We found that information accessibility affected decisions to seek information, but not self-reported curiosity. This suggests that curiosity is unhindered by the practical costs of information search.

Ease of Access to Information Does Not Impact Curiosity

Groupitizing, the ability to use perceptual grouping to facilitate enumeration, appears with schooling and predicts math achievement (Guillaume et al., 2023). Our study (Spring 2025) investigates the cognitive mechanisms underlying its development and enhancement during early education.

We propose that groupitizing among young children involves two core processes: (i) subitizing to identify the numerosity of small subgroups and (ii) using perceptual grouping to perform mental arithmetic, such as addition or multiplication.

To test this hypothesis, children from kindergarten through second grade will perform an enumeration task involving dot sets organized into various grouping structures (subgroups with equal or differing numerosity) and various grouping patterns (subgroups with consistent or varied shapes). Response times and error patterns will be analyzed, alongside children's counting knowledge, arithmetic skills and math achievement.

Our results aim to elucidate the cognitive mechanisms underlying groupitizing across development, and the links between foundational perceptual skills and higher-order mathematical reasoning.

Downloads

Next from CogSci 2025

A Neurosymbolic Model of Human Reasoning on the Abstraction and Reasoning Corpus

Similar lecture

The relationship between statistical learning and different facets of language ability: evidence from auditory and visual modalities

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES

.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

A Neurosymbolic Model of Human Reasoning on the Abstraction and Reasoning Corpus

Similar lecture

The relationship between statistical learning and different facets of language ability: evidence from auditory and visual modalities

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES

Downloads