United States

The relationship between symbolic and nonsymbolic representation has been a subject of long-standing debate, with the common system model and the separate systems model providing contrasting predictions. To test these models, we asked participants to compare the size of animals or numerical value of numbers, presented in symbolic-symbolic, symbolic-nonsymbolic, and nonsymbolic-nonsymbolic formats. Consistent with the common system model, performance improved as the ratio between stimuli increased in both animal and number domains, regardless of the format. However, supporting the separate systems model, we observed a switch cost in the symbolic-nonsymbolic number comparison task. Future research should explore the factors contributing to these mixed findings across domains.

CogSci 2025

Mixing Words and Pictures: Mixed Evidence for Common Conceptual Representations

concepts and categories

psychology

representation

### 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

Information theory characterizes how signals are optimized for transmission from source to receiver across noisy channels, yet little is known about how these principles manifest when the receiver's capabilities change over time. Using child-directed speech as a natural experiment, we analyzed >7.5 million phones in North American English caregiver speech to children aged 3-44 months (N=218) from the CHILDES database. We found that while the relative frequency of individual phones remained stable over this developmental time period, phonological informativity increased from early infancy (3-8 months) through toddlerhood (27-32 months), before plateauing in the preschool years. This result suggests that speech directed to children sounds less redundant, with a phonological structure that is harder to predict in context, as children progress through early childhood. Our findings demonstrate how linguistic signals may be optimized to accommodate receiver (child) characteristics, with implications for both general principles of information transmission and theories of how children carve out linguistic representations and patterns from limited, noisy input.

Frequency and informativity of phonological input directed to children in the first four years of life

Inductive concept learning requires making inferences about target categories based on specific examples. Two factors which influence this process are type of learning task and the nature of the items available for comparison. However, the literature remains inconsistent on which combination of factors best facilitates concept learning. Moreover, much of the present literature focuses on artificial categories with arbitrary boundaries, leaving open the question of how best to improve learning for natural categories. We report two experiments on natural category learning, which cross learning mode (classification vs observation) with comparison type (match vs. contrast vs. control). Across both experiments, we find evidence of an observation advantage and some evidence for a contrast advantage (Experiment 1). These findings offer evidence against a classification advantage during natural category learning, which some studies have shown, and highlight the critical need for investigating the factors that impact the efficacy of classification and observation learning.

Classification Versus Observation through Within- and Between-Category Comparison

The gradual nature of historical language change is widely acknowledged. We explore a syntactic change model that offers a new view into the theoretical difference between classical and neural network claims about language encoding. Most prior treatments of grammaticalization fail to account for how, exactly, new forms arise, focusing instead on change following innovation. A phenomenon relevant to the innovation puzzle is Quantitative Anticipation of Qualitative Change in Grammaticalization (QAQCG)---gradual statistical changes anticipate structural changes. Although prior researchers have given phenomenological descriptions, we know of no rigorous method for testing whether QAQCG exists. Here, we quantitatively examine the case of English "a lot" which has grammaticalized an Adverb function from a Noun Phrase function. A simple feedforward neural network implements QAQCG, predicting a curving trajectory in probability space. Bayes Factor analysis supports the network over a classically-motivated linear model, highlighting continuity and nonlinearity as distinctive theoretical claims.

Quantitative Qualitative Correspondence in Grammaticalization

Modern machine learning systems have demonstrated substantial abilities with methods that either embrace or ignore human-provided knowledge, but combining benefits of both styles remains a challenge.  One particular challenge involves designing learning systems that exhibit built-in responses to specific abstract stimulus patterns, yet are still plastic enough to be agnostic about the modality and exact form of their inputs.  In this paper, we investigate what we call The Ungrounded Alignment Problem, which asks How can we build in predefined knowledge in a system where we don't know how a given stimulus will be grounded?  This paper examines a simplified version of the general problem, where an unsupervised learner is presented with a sequence of images for the characters in a text corpus, and this learner is later evaluated on its ability to recognize specific (possibly rare) sequential patterns.  Importantly, the learner is given no labels during learning or evaluation, but must map images from an unknown font or permutation to its correct class label.  That is, at no point is our learner given labeled images, where an image vector is explicitly associated with a class label.  Despite ample work in unsupervised and self-supervised loss functions, all current methods require a labeled fine-tuning phase to map the learned representations to correct classes.  Finding this mapping in the absence of labels may seem a fool's errand, but our main result resolves this seeming paradox.  We show that leveraging only letter bigram frequencies is sufficient for an unsupervised learner both to reliably associate images to class labels and to reliably identify trigger words in the sequence of inputs.  More generally, this method suggests an approach for encoding specific desired innate behaviour in modality-agnostic models.

The Ungrounded Alignment Problem

Are the ‘portioning’ readings (‘several beers’) and ‘grinding’ readings (‘a bit of pear’) the result of lexical derivations with real-time processing effects? The evidence is inconclusive. While Frisson and Frazier (2005) argue that these readings are visible as cost, Lima (2019) reports no such effects. We address this inconsistency through two English self-paced reading experiments. Experiment I testing the ‘portioning’ reading (‘several pears’ vs. ‘several beers’) reveals no additional processing effects. By contrast, Experiment II testing the ‘grinding’ reading (‘a bit of beer’ vs. ‘a bit of pear’) reveals higher reading times for the ‘grinding’ condition one word after the critical noun. These results provide empirical support for an asymmetrical relationship between individuated ‘count' and non-individuated ‘mass' readings. Whereby the former represent the default conceptual representation resulting in no cost, the latter result from a conceptual expansion to include a container-containee conceptualization, which is done in real-time, resulting in cost.

Reexamining Mass/Count Flexibility in the Nominal Domain: A Real-Time Comprehension Study

Ensuring that all children have access to evidence-based reading instruction requires scalable solutions that respect linguistic diversity (Castles et al., 2018). Kalulu is a fully automated system designed to develop phonics-based reading programs for any symbol-to-sound language. Grounded in cognitive science, it generates instructional materials—including books, paper-based games, and digital applications—tailored to the phonological structure of each language. Leveraging AI, Kalulu analyzes grapheme–phoneme correspondences to build a progression of two mappings per week, enabling 100% decodable texts while integrating reading, writing, and vocabulary instruction.

Initially tested in France with over 1,000 children, Kalulu is now being deployed in Brazil (1,000+), Colombia (500+), and Mayotte (300+), with ongoing expansion into Argentina. This poster outlines the automation pipeline, field implementation strategies, and current cross-linguistic research in collaboration with local school districts and NGOs. We aim to show how cognitive science can accelerate the global scaling of evidence-based literacy instruction—providing free and open-access learning.

Kalulu: Evidence-Based Adapted Phonics Instruction for Literacy Across Languages

Narrative interaction plays an important role in shaping people's beliefs and behaviors both online and in the offline world. We present an experiment examining whether a simple intervention of effect prompting---asking participants to list the effects of complex events---impacts the narrative framing of their networked interactions. After reading a text-based narrative about the Fukushima nuclear disaster, participants in a fully connected network interacted with their neighbors and received rewards for submitting hashtags that matched those of their network partners. Half of the groups received an \textit{effect-prompting} intervention, which shifted participants toward producing more effect-oriented hashtags during networked interactions. We found that the effect-prompting instruction influenced the hashtags participants generated during the network interaction. However, the extent of this shift in hashtags depended on how likely the group was to achieve global coherence.We also examined these dynamics with networks of interacting large language model (LLM) agents using Llama-3.1-8B-Instruct. The study highlights how language-based prompting can subtly shift the narrative framing of online communication.

Effect-prompting shifts the narrative framing of networked interactions

The present study examined how participants (N = 298) assessed causality, blameworthiness, foreseeability, and counterfactuality of an AI or human therapist, across three levels of empathy, in comparison to their supervisor and a recommending clinician. We found that participants judged the human therapist as more causal and blameworthy than their supervisor when medium or low empathy levels were displayed, whereas no difference emerged between the judgments of the AI therapist and its supervisor across all of the empathy levels. Additionally, participants did not differentiate causality and blameworthiness between the AI and human therapists, regardless of the empathy level. However, they did perceive the human therapist as foreseeing the outcome more than the AI therapist in the medium and low empathy levels. Qualitative analysis revealed that participants considered the directness of the causes to the outcome, counterfactual reasoning, and inherent limitations of AI when making judgments.

Cause and Blame Attribution to AI and Human Agents in Mental Health Context

Food choices, as a type of value-based decision, are affected by environmental cues. We conducted three studies to investigate the computational mechanisms through which background music influences the food choice-making process. Hierarchical drift-diffusion modeling revealed that nature-related music led to higher drift rates than urban-related music, indicating faster evidence accumulation toward certain choices. Specifically, participants processed the value of vegetable-forward meals more efficiently when exposed to nature-related music compared to urban-related music. Moreover, the effect of music on the drift rate varied with the vividness of music-induced mental imagery and the perceived identity of performers (human or robot). Collectively, these findings reveal the computational mechanisms underlying the influence of environmental cues on value-based decision-making.

The Computational Mechanism of How Music Influences Food Choices:  A Drift-Diffusion Model Analysis

Whether language is essential, sufficient, or a tool for numerical cognition has been hotly debated. Here, we investigate the influence of language on quantity representations by comparing embeddings from vision-only Transformer models (ViTs) and vision-language models (VLMs) exposed to image pairs depicting either the same or different stimulus quantities. If linguistic exposure stabilises quantity representations, VLMs should produce more distinct representations for image pairs with differing numerosity and more similar representations for those with identical numerosity than ViTs. We operationalized this as the variance in Cosine Similarity in response to either categorical (same/different) or continuous differences in stimulus numerosity. We find that VLMs and ViTs are sensitive to the numerosity of visual stimuli, that this sensitivity increases with layer depth, and that VLMs exhibit slightly more sensitivity to image numerosity than ViTs. This work provides initial support for the claim that linguistic exposure can, in principle, stabilise quantity representations.

Downloads

Next from CogSci 2025

Frequency and informativity of phonological input directed to children in the first four years of life

Similar lecture

Symbolic numerical generalization through representational alignment

.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

Frequency and informativity of phonological input directed to children in the first four years of life

Similar lecture

Symbolic numerical generalization through representational alignment

Downloads