United States

This research introduces an intelligent mentoring system that utilizes behavioral logs and cognitive modeling to provide personalized learning interventions. The system analyzes semantic patterns during web browsing to estimate cognitive parameters and predict task completion times. By employing CLIP, the system evaluates semantic alignment between learning objectives and accessed resources. Our experimental study with eight postgraduate students serves as a starting point for ACT-R cognitive parameter estimation. The system tracks both content relevance and user interest as internal human dynamics to maintain engagement with study materials. By integrating ACT-R architecture to estimate working memory capacity, attention span, and anxiety levels, the system maps these parameters to observed browsing behaviors. This dual-stage approach enables real-time cognitive estimation and personalized schedule generation, establishing a foundation for adaptive learning technologies that provide tailored interventions based on semantic dynamics analysis and cognitive parameter estimation.

CogSci 2025

Developing a Mentoring System Based on Behavior Logging and Personalized Cognitive Modeling

quantitative behavior

computational modeling

learning

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

Letter-position encoding is one of the constituent processes in
visual word recognition. While the existing models attempt to
explain letter-position encoding in English & other European
languages written using the Roman Script, letter-position
encoding in Hindi written using the Devanagari Script has not
been studied in detail. Given that Hindi is spoken/read by over
520 million people in India and the unique properties of the
Devnagari Script (Vaid & Gupta, 2002; Kandhadai & Sproat,
2010; Share et al., 2015), the current study sought to investigate
letter-position encoding in Hindi. 66 participants performed a
lexical decision task that employed six-prime conditions to
compare hypotheses from a) the position-specific coding
scheme, b) local context sensitive coding scheme and the c)
position overlap coding scheme. Interestingly, the results
showed that none of the aforementioned coding schemes could
satisfactorily explain the obtained data. These findings may be
used to question the generalizability of the extant letter-
position encoding schemes to the relatively understudied
languages such as Hindi, which use scripts different from the
Roman script.

Existing Models May Not be Able to Explain Letter-Position Encoding in Hindi: Evidence from a Priming Study

Real-time human-artificial intelligence (AI) collaboration is crucial yet challenging, especially when AI agents must adapt to diverse and unseen human behaviors in dynamic scenarios. Existing large language model (LLM) agents often fail to accurately model the complex human mental characteristics such as domain intentions, especially in the absence of direct communication. To address this limitation, we propose a novel dual process multi-scale theory of mind (DPMT) framework, drawing inspiration from cognitive science's dual process theory. Our DPMT framework incorporates a multi-scale theory of mind (ToM) module to facilitate robust human partner modeling through mental characteristic reasoning. Experimental results demonstrate that DPMT significantly enhances human-AI collaboration, and ablation studies further validate the contributions of our multi-scale ToM in the slow system.

DPMT: Dual Process Multi-scale Theory of Mind Framework for Real-time Human-AI Collaboration

This study investigates the structural alignment of conceptual representations accessed from visual and linguistic modalities in 8-9-year-old children and adults. Using a Spatial Arrangement (SpAM) task, participants organized familiar items from two categories – household items (HHI) and vegetables & fruits (VF) – presented separately as images and written words on a two-dimensional grid. Representational dissimilarity matrices were computed based on item distances within each modality and analyzed for structural alignment across modalities. Results showed significantly lower cross-modal alignment in children compared to adults, suggesting ongoing developmental changes in the structure of conceptual representations. Additionally, cross-modal alignment was higher for VF than for HHI categories in both age groups, indicating category-specific variations in conceptual organization and its refinement. These findings provide insights into the gradual refinement of the structure of the lexical-conceptual system, extending beyond item-level lexical learning.

Structural Alignment Across Visual and Linguistic Modalities:  A Developmental Refinement Perspective

Generative artificial intelligence (GenAI) is rapidly advancing, providing innovative tools and methods for a wide range of applications. Among these, Generative Agents, a key domain in GenAI, are valued for their fine-grained definitions and simulations of human-like behaviors. These agents provide new avenues for studying and modeling various domains, including social interactions, education, and cognitive sciences. However, existing works suffer from cognitive dynamics disconnect and affective absence, which hinder researchers from exploring human-related cognitive processes in depth. To address these limitations, we propose EvoAgents, a novel cognitive-driven framework that pioneers the exploration of dynamic personality evolution in Generative Agents. By defining the emotional content of agents and integrating a cyclical personality evolution cycle, EvoAgents represent a significant step toward creating more adaptive and authentic agent behaviors. Comprehensive simulations and evaluations show that EvoAgents achieve superior performance in key automated metrics compared to prior work, while uniquely enabling reasonable and robust personality evolution processes that align with cognitive and psychological expectations. By constructing a new simulation environment, SmallClassroom, based on the EvoAgents framework, we validate the framework's ability to provide deeper cognitive insights into social dynamics, aligning closely with established psychological theories.

EvoAgents: A Cognitive-Driven Framework for Personality Evolution in Generative Agent Society

Automated vehicle (AV) validation faces the "billions of miles" challenge, requiring high-fidelity simulations to replicate diverse interactive driving behaviors for safety. Traditional methods oversimplify by using uniform behavioral models, ignoring the diversity of human driving styles, which are deeply influenced by individual psychological traits. This research introduces a condensed framework for representing interactive driving styles, by incorporating these psychological dimensions, balancing completeness and complexity. Key features include: i) individual style recognition via attention mechanisms and hierarchical contrastive learning, capturing subtle cognitive-based interaction patterns that reflect underlying differences in driver psychology (e.g., risk tolerance, decision-making heuristics); ii) scenario-independent style compression, filtering external factors to extract intrinsic driver intentions; iii) dimensionality-aware refinement, mapping complex behaviors to low-dimensional psychological axes for efficient computation. Tests on the NGSIM dataset reduced testing complexity by decoupling styles from scenarios. Compared to traditional methods, style distinctiveness improves by 28% (entropy-based), with 85% edge-case behavior coverage. This framework supports scalable AV testing by integrating diverse, psychologically-informed driving styles without combinatorial complexity.

Condensed Representation Learning for Interactive Driving Styles Recognition

Based on research in cognitive psychology, humans typically seek a balance between their preference for familiar things and the exploration of new ones during decision-making. Therefore, studying the relationship between accuracy and diversity in recommendation systems is particularly meaningful. In recent years, recommendation systems based on Graph Neural Networks (GNNs) have garnered significant attention for enhancing recommendation accuracy or diversity. However, existing works often improve accuracy or diversity at the expense of the other aspect, which is inconsistent with the complex needs of users. In this paper, we propose a novel Recommendation model that Balances Diversity and Accuracy with GNNs, called BDARec. Firstly, BDARec proposes a balanced neighborhood aggregation strategy to select diverse and accurate neighbor nodes for updating node embeddings in user-item bipartite heterogeneous graph. Secondly, to accelerate the convergence of BDARec, an enhanced category-boosted negative sampling strategy is proposed to select negative samples from the same category positive samples with a certain probability. Thirdly, we put forward a dynamic feature for each item to measure the importance of items in training phase. Finally, we conduct extensive experiments on three real-world datasets. Experimental results show that our model can even improve recall by 22.04%, hit_ratio by 16.46%, and coverage by 10.27% when compared to the state-of-the-art comparison algorithm, which verifies that the proposed model can achieve the best balance between diversity and accuracy.

BDARec: Balancing Diversity and Accuracy of Recommendation Model with Graph Neural Networks

Intelligence cognition, as a pivotal element in intelligence activities, encompasses a series of psychological processes, cognitive operations, and behavioral manifestations involved in the acquisition, processing, analysis, and utilization of intelligence throughout the entire process. This study reviews the current status of intelligence cognition research, explores the theoretical foundations of intelligence cognition， elucidates the cognitive components embedded in the intelligence research workflow, and identifies key issues that need to be addressed in the future field of intelligence cognition.

Research on Fundamental Issues of Intelligence Cognition

Session-based recommendation aims to capture user intent from short-term, anonymous interaction sequences to recommend relevant items. From a cognitive science perspective, understanding user intent is closely tied to how humans process information, allocate attention, and make decisions under limited cognitive resources. While existing session-based methods mainly rely on ID-based modeling, such approaches face severe data sparsity and lack alignment with how users cognitively process information. Incorporating modality information can alleviate this issue. However, simple integration of ID and multimodal information often results in modality underfitting, limiting the effective use of multimodal features. To address these challenges, we propose SRIM(Session-based Recommendation with ID and Modality), a model that integrates ID and multimodal representations through a two-phase strategy: independent training followed by joint optimization. SRIM can better capture session-level intent by simulating users' actual perceptual contexts. Experiments on three real-world datasets demonstrate that SRIM significantly outperforms existing methods in session recommendation. The code for SRIM is available on GitHub {https://github.com/liang-tian-tian/SRIM}.

Capturing User Intent through Integration of Item ID and Modality Information in Session Recommendation

Multimodal sentiment analysis (MSA) recognizes human sentiments with various data modalities. Existing works primarily focus on efficient feature extractors for each modality or multimodal fusion frameworks. However, they do not exploit the sentiment-related prior knowledge in the human brain, limited in their performance when sentiment cues tend to be more implicit and ambiguous. To address this problem, we propose a rationale-guided prompt learning optimization framework (RaPo) inspired by the sentiment chains of biological brains. Specifically, we adopt a chain-of-thought prompt to analyze images with a large visual-language model, generate the corresponding contextual captions and rationales, which are then combined to derive the sentiment prompt. Finally, the prompt is used to optimize the RaPo framework through the designed rationale-guided prompt tuning. Experiments on several MSA tasks consistently show an outperformance of several state-of-the-art methods, with an average increase in accuracy by 2.8%.

A Brain-Inspired Multimodal Sentiment Analysis Framework via Rationale-Guided Representation

Previous research has shown that participants exhibit better memory performance when consciously closing eyes (EC) compared to keeping eyes open (EO). However, the underlying dynamical mechanism remains unclear. Here, we propose a reservoir computing (RC) algorithm based on EEG-derived  functional connectivity of  brain networks in resting state to simulate the brain’s memory processes and reproduce the EC-related memory advantage. Our findings indicate that, compared to EO-based connectivity, the RC constructed from EC-based resting-state connectivity demonstrates superior memory performance. Further graph-theoretical analysis reveals that the EC networks exhibit stronger modularity and the modularity index are positively correlated with  memory ability. Therefore, we conclude that the functional connectivity of whole brain underlies memory function and  the broad reorganization of connectivity in the EC state leads to its memory advantage over the EO state.

Downloads

Next from CogSci 2025

Existing Models May Not be Able to Explain Letter-Position Encoding in Hindi: Evidence from a Priming Study

Similar lecture

Emotional Parameters in Cognitive Architecture: Examination Through Simple Memory Performance

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES