United States

CogSci 2025

Functional category induction with theory-neutral cognitive biases

computational modeling

computer science

language acquisition

machine learning

syntax

## Welcome!
"I am excited to welcome you to this year’s edition of the Conference on Empirical Methods in Natural Language Processing! Importantly, it marks the 30th edition of EMNLP. With over 8,000 submissions, more than 3,000 accepted papers, and thousands of attendees, we have come a long way from that first
workshop, which had 14 accepted papers. As the field looks ahead, Suzhou is the fitting location for celebrating this milestone: rooted in a long literary tradition, yet modern and forward-looking, and home to a large share of the NLP community."<br>

*Message from the General Chair, Dirk Hovy*





<br>

EMNLP 2025

Celebrate 30 Years of EMNLP! 
EMNLP 2025 will be held in Suzhou, China from November 5th to November 9th, 2025.

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.

Manipulating Predictive Focus Improves the Taste Appreciation of Coffee

The approximate number system (ANS) enables humans to rapidly estimate numerical quantities without relying on counting, yet the computational processes underlying the ANS remain mysterious. Here, we propose a mechanistic model for the ANS, built on the core idea that when presented with a stimulus array, observers first form a template representation through ensemble averaging and subsequently engage in a template-matching process, whereby each array item is compared to the template. With a limited set of plausible assumptions about the inherently noisy computational processes, our model naturally accounts for Weber’s law, the main hallmark of number psychophysics, systematic numerosity underestimation and the coherence illusion. We further test two novel predictions about stimulus factors that modulate the strength of the coherence illusion and demonstrate high fidelity between predicted and obtained data. This model offers new insight into the computational mechanisms by which ANS representations are derived from visual input.

Noisy template matching: A mechanistic model of approximate number perception

Research in experimental syntax typically assumes that the five-point Likert scale offers an ordinal probe that maps monotonically onto a latent degree of sentence acceptability. We challenge that assumption by showing that, when the stimulus space is densely populated, speakers repurpose the scale into an anchor-and-baseline device. Two large-N experiments (Russian and Serbo-Croatian; N=237; 120 permutational word-order variants per language) elicited over 28000 sentence acceptability ratings. Plotting Shannon entropy of the response distribution against the mean rating reveals a robust ’entropy arch’: uncertainty climbs to a sharp peak at the midpoint and collapses toward both ends. We interpret the arch as the quantitative fingerprint of constraint competition: the scale extremes serve as categorical anchors (’completely acceptable’ vs. ’completely unacceptable’), while the center functions as a floating baseline against which speakers register maximally uncertain, cue-balanced configurations for which grammatical, information-structural and frequency-based cues pull in opposite directions. Our findings reframe Likert data as the outcome of dynamic calibration rather than static preference strength and provide a simple diagnostic, entropy profiling, for locating linguistic ’tipping-point’ constructions. Beyond sentence acceptability, the approach offers a principled way to map regions of maximal competition in any domain where categorical anchors and graded uncertainty coexist.

Dense Sentence Sets Induce an Anchor-and-Baseline Strategy in Likert Scale Acceptability Judgments

It has been argued that when scientists employ algorithmic tools to assist in problem-solving, epistemic monocultures may emerge in which research tools, topics, findings, etc. are homogenized. As a result, fertile areas of research might be left unexplored, impeding scientific progress. To explore the nature of these epistemic monocultures, we develop an agent-based model where agents have the ability to query an AI system to assist in their search of an epistemic (NK) landscape. In general, we find that AI use negatively affects the community of researchers by reducing heterogeneity, but both the rate of AI queries and how AI is used impact the ultimate success of the community. We then implement a potential solution suggested in the literature, AI personalization, and find somewhat mixed results on its potential for mitigating homogenization in research communities.

Epistemic Monocultures and the Effect of AI Personalization

Do children believe AI deserves to be treated morally as humans, and if so, why? In this study, eighty children aged four to eight were randomly assigned to interact with a human or AI partner in a naturalistic storytelling activity, after which their moral considerations and rationale were elicited. The results indicated that although children believed they should treat both AI and humans morally, the motives behind their beliefs differed. Children's moral considerations for humans were driven by empathy for others' needs and feelings, whereas their considerations for AI were motivated by the desire to preserve its functional utility. These motives were corroborated by children's denial of AI’s capacity to have feelings or basic physical needs. This study contributes to the literature on children's moral development by demonstrating that while their moral reasoning extends to non-human entities, their justifications reflect a distinct, domain-specific understanding rather than an anthropocentric confusion.

Empathy or Utility: Children's Reasoning for Moral Consideration of AI

Literature indicates that lingua franca use in clinical settings impacts bilinguals communicating in their second language (L2)—but what drives these changes? This study examined how language, proficiency, and cognitive load influence emotionality during traumatic autobiographical memory (AM) retrieval. Building on evidence that L2 use diminishes emotional resonance, we hypothesized that proficiency would moderate this effect. Additionally, we expected lower emotionality in first-language (L1) recall under cognitively demanding conditions, mirroring second-language effects. Greek native speakers (N=107) recalled a traumatic AM in L1, L2 (English), or L1 under cognitive load (visuospatial working memory task) while undergoing physiological EDA assessment. Self-reported arousal, distress, and traumatic symptoms were measured. Notably, L2 alone did not lower emotional intensity—only lower L2 proficiency did, highlighting its key moderating role. Cognitive load during L1 retrieval significantly decreased emotionality, supporting resource-dependent emotional processing models. These findings reveal proficiency and cognitive demands’ interplay in shaping bilingual emotional processing.

Same memory, different words: The Role of Language, Proficiency and Cognitive Load in Traumatic Autobiographical Memory Recall

Stimulating language learners’ engagement is essential to successful second language acquisition, but it can be hard to translate this intuition into effective learning resources. In the first large scale investigation into the linguistic and pragmatic features that make an educational conversation interesting, we collected interest ratings for 64 conversations between teachers and second language learners of English. We provide proof of concept that - despite the high degree of subjectivity involved in perceptions of interest - it is possible to extract features that make a conversation interesting for the average learner. Specifically, concreteness, comprehensibility, and uptake (i.e., the degree to which a teacher and a student’s turn build on one another) all had unique relations to interest in our data. These findings lay the foundations for future work on the optimization of AI tutors for more engaging language learning interactions.

What makes a conversation interesting? Linguistic features predictive of interest in educational conversations between teachers and learners of English

The current study examines 120 5 to 10-year-old children’s beliefs about adults’ abilities to use and fix tablet technology when those adults belong to varying gender (man, woman) and age (young, old) categories. The results indicate that, overall, children appear to prioritize age over gender when judging adults’ technological knowledge, with children choosing younger adults as more competent at using and fixing tablets than older adults. In addition, when evaluating adults of the same age category (e.g., a young man and a young woman), children show in-group gender-based preferences where girls choose women and boys choose men. This in-group preference is more pronounced in children’s selections of adults when determining who would be better at fixing tablets than who would be better at using these devices. Implications for children’s developing ability to consider intersectional identities based on gender and age, and for their STEM learning, are discussed.

Children Prioritize Age over Gender when Evaluating Adults’ Technological Knowledge

The low prevalence effect (LPE), a decrease in target detection performance as target prevalence decreases, is a concern in real-world visual search tasks, such as baggage screening. Unfortunately, much of the research into the LPE and its potential countermeasures may not represent the challenges of other real-world search tasks, such as sonar and security monitoring, in which objects in the search environment exhibit movement. Additionally, target (e.g., submarine) and non-target (e.g., merchant ship) movement may interact with target prevalence to further decrease target detection performance, but these factors have not been systematically manipulated to determine their effects. In Experiment 1, high and low prevalence targets occurred in search conditions with static or randomly-moving objects. Although there was no significant interaction effect detected between target prevalence and motion, these conditions independently contributed to significantly lower hit rates. In Experiment 2, a higher prevalence target was included as an attempted LPE countermeasure. Observers searched for a relatively high and a low prevalence target in the moving search task. The addition of a higher prevalence target improved target detection overall in a search environment with moving objects, serving as a possible countermeasure to the LPE.

The Low Prevalence Effect with Random Motion Stimuli in a Visual Search Task

Visual working memory (VWM) plays a fundamental role in cognitive processes, such as perception, attention, and reasoning. However, existing approaches to modelling VWM are not integrated into cognitive architectures and lack interpretability with respect to their parameters. To address this limitation, we propose a novel VWM model based on the well-established Semantic Pointer Architecture (SPA). In contrast to previous works, our model is the first to integrate a VWM model with a cognitive attention model. It only requires three interpretable hyper-parameters: spatial capacity, feature certainty, and memory decay. We experimentally show that our base model without memory decay replicates the set-size effect and swap errors of human data on a continuous reproduction task. More importantly, we show that by introducing a memory decay, we can achieve a statistically significant (p ≪ 0.001) improvement in model fit, suggesting a potentially important role of memory decay in VWM. Further, our VWM model can be easily extended to model pre- and post-cue conditions, consistently achieving KL divergence between modelled and human performance of less than 0.05.

Downloads

Next from CogSci 2025

Manipulating Predictive Focus Improves the Taste Appreciation of Coffee

Similar lecture

A Computational Simulation of Language Production in First Language Acquisition