United States

Manga has gained global popularity, yet how its visual elements, such as characters, text, and panel layouts, reflect the uniqueness of individual works remains underexplored. This study investigates the contribution of panel layouts to manga identity through both quantitative and qualitative analysis. We trained a deep learning model to classify manga titles based solely on facing page images, and performed ablation experiments by removing characters and text, retaining only panel frame structures. Using 10,122 images from 104 works in 12 genres in the Manga109 dataset, we demonstrate that panel layouts alone enable high-accuracy classification. Grad-CAM visualizations further reveal that the models focus on layout features such as size, spacing, and alignment. These findings suggest that panel layouts encode work-specific stylistic patterns and support visual narrative comprehension, highlighting their role as a key component of manga&#39;s visual identity.

CogSci 2025

How Panel Layouts Define Manga: Insights from Visual Ablation Experiments

art and cognition

computer-based experiment

computer science

creativity

machine learning

The purpose of the AAAI conference is to promote research in artificial intelligence (AI) and scientific exchange among AI researchers, practitioners, scientists, and engineers in affiliated disciplines. 

AAAI 2021

Manga has gained global popularity, yet how its visual elements, such as characters, text, and panel layouts, reflect the uniqueness of individual works remains underexplored. This study investigates the contribution of panel layouts to manga identity through both quantitative and qualitative analysis. We trained a deep learning model to classify manga titles based solely on facing page images, and performed ablation experiments by removing characters and text, retaining only panel frame structures. Using 10,122 images from 104 works in 12 genres in the Manga109 dataset, we demonstrate that panel layouts alone enable high-accuracy classification. Grad-CAM visualizations further reveal that the models focus on layout features such as size, spacing, and alignment. These findings suggest that panel layouts encode work-specific stylistic patterns and support visual narrative comprehension, highlighting their role as a key component of manga's visual identity.

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.

This study explores the psychological impact of AI-generated and user-manipulated images on body image perception, particularly in the context of social media platforms. Focusing on young adults, the research examines their ability to identify A.I.-enhanced, user-enhanced, and unaltered images. Results indicate that participants can readily detect AI-enhanced images due to exaggerated features but struggle to identify subtle alterations from traditional photo-editing apps. Interestingly, participants showed a preference for minimally edited or unaltered images, despite faster detection of AI- enhanced images. Qualitative data suggest a divide in participants' attitudes toward AI manipulation: some expressed concern about its effects on body image and self-esteem, while others expressed indifference. These findings highlight the increasing difficulty in distinguishing authentic content from digital manipulation and raise important questions about rapidly evolving definitions of beauty and authenticity. Overall, findings underscore the need for media literacy interventions to address these challenges.

Perceptions of A.I.-Enhanced Bodies: Autonomy, Authenticity, and Preferences Among Young Adults

Social stimulus often triggers a stress response, influencing learning performance depending on the task’s nature. Previous research shows that social pressure before learning impairs performance in rule-based (RB) learning tasks, relying on hypothesis-testing mechanisms. Conversely, it may enhance performance in information-integration (II) learning tasks, which depend on procedural memory systems.
This study examined the effects of social feedback stimuli (e.g., smiling or angry faces), compared to non-social feedback stimuli (e.g., green circles or red crosses) on performance in RB and II tasks. Participants were assigned to either an RB or II task and received either social or non-social feedback during the task.
Results showed that social feedback significantly improved performance in the II task compared to non-social feedback, but did not enhance RB task performance. These results shed light on the distinct cognitive mechanism underlying categorical learning and emphasize the importance of feedback design in educational and training environments.

Impact of Social Feedback Stimulus on Information-Integration Category and Rule-based Category Learning

Prospect Theoretic Loss aversion (λ) suggests we overvalue losses compared to gains during decision making but the influence of age on loss aversion is unclear. We conducted a random-effects meta-analysis on 7 studies that measured loss aversion and the age of participants. The pooled estimate (β=0.0886, 95%CI [0.0087 0.1686] revealed an effect of age on loss aversion. Considering age as continuous (n=1103, 5 studies; β = -0.005233, t= -2.511, p = 0.0122) or categorical (n=1856, 8 studies; β=0.00856, t= 0.111, p = 0.912) reflected substantial differences. Dividing participants broadly (< 46 and > 45 years) to explore differences in estimates of loss aversion (λ>1, λ =1, λ<1), also found significant differences, X2 (2, 1856)=7.5362, p = 0.0231. These findings underscore the veracity of meta-analysis for advancing and explicating previously underexplored aspects of decision theory.

A Meta-analysis of Age-related effect on Loss Aversion

Automatic emotion recognition using electroencephalography (EEG) signals has garnered significant attention in recent years. While multi-source domain adaptation methods provide a promising framework for cross-subject emotion recognition, the distributional discrepancies among different source domains often result in negative transfer. To address these challenges, we propose a two-stage Dynamic Domain Selection and Pseudo-label Refinement (DDSPR) model. In the first stage, we introduce a novel Dynamic Domain Selection (DDS) module and an Agent Domain Adaptation Strategy (ADAS) to dynamically select and align source domains. In the second stage, a confidence-based pseudo-label correction strategy is employed to refine target domain labels and mitigate noise. We evaluate the proposed model through cross-subject experiments on the SEED and SEED-IV datasets, achieving accuracies of 91.50% ± 7.05 and 78.05% ± 13.56, respectively.The results demonstrate its effectiveness in emotion recognition performance.

DDSPR: Dynamic Domain Selection and Pseudo-label Refinement for Cross-Subject EEG-based Emotion Recognition

Communication is essential for enhancing the cognition and cooperation of agents in multi-agent reinforcement learning (MARL). However, existing methods often rely on predefined and rigid cognitive patterns, which cannot adapt to dynamic environmental changes and complex inter-agent interactions. In this work, we introduce the Hierarchical Cognitive Graph Autoencoder (HCGA), an adaptive framework that addresses these limitations. HCGA represents inter-agent messages as nodes in a graph with learnable edges, employs a grouping mechanism to integrate related local information into compact latent representations, and then applies hierarchical aggregation to construct a comprehensive global cognition. This approach effectively distills essential information and adaptively uncovers cognitive patterns from dynamic environments, thereby enhancing the overall robustness and efficiency of cognitive processing in MARL tasks. Experimental results demonstrate that HCGA significantly outperforms state-of-the-art methods across various MARL tasks, highlighting its robustness, adaptability, and efficiency.

Hierarchical Cognitive Graph Autoencoder for Multi-Agent Reinforcement Learning

Large Language Models (LLMs) have been extensively employed in Legal Judgment Prediction (LJP) in recent years. However, existing LLM-based methods often fail to effectively simulate the cognitive processes of human judges, particularly in keyword extraction, leading to suboptimal predictions. Inspired by cognitive science, we propose KnowJudge, a knowledge-driven framework, which explicitly models the cognitive process of legal decision-making, leveraging keyword extraction and precedent-based enhancement to guide LLMs in structured legal reasoning. By integrating external legal knowledge tailored to fact descriptions, it refines keyword identification and selects relevant case precedents, thereby mitigating ambiguity in legal judgment.
Unlike conventional methods that rely on fine-tuning, KnowJudge improves performance purely through cognitive-process simulation. Experiments on five benchmarks show that KnowJudge outperforms baseline methods, including both general and legal LLMs.

KnowJudge: A Knowledge-Driven Framework for Legal Judgment Prediction

Accurate recognition of human emotions from EEG signals plays a critical role in affective computing and human-computer interaction. However, existing methods face significant challenges in effectively capturing the sparse, dynamic, and energy-dependent characteristics of brain activity during emotional experiences. To address these challenges, we propose a novel framework, Spatial-Energy-Aware Dynamic Filtering with Sparse Graph Convolutions (SEASGC), which rethinks EEG graph modeling from three perspectives: (1) sparse graph construction to adaptively capture the essential functional relationships between brain regions, (2) dynamic and location-dependent filtering to model nonlinear interactions between EEG nodes, and (3) energy-aware feature aggregation to leverage energy changes as critical indicators of emotional intensity. By explicitly integrating these principles, SEASGC provides a more comprehensive representation of EEG signals for emotion recognition. Extensive experiments on benchmark EEG emotion datasets demonstrate that SEASGC achieves state-of-the-art performance, highlighting its effectiveness and generalizability in modeling the complex spatial-spectral dynamics of EEG signals.

Spatial-Energy-Aware Dynamic Filtering with Sparse Graph Convolutions for EEG Emotion Recognition

The proliferation of DeepFake has engendered widespread societal concerns, positioning its detection as a pressing imperative. Although existing studies have utilized single-subject EEG to distinguish between real and AI-generated content (AIGC), there is still a lack of research exploring dual-brain EEG and multimodal experimental paradigms. This study introduced a novel experimental paradigm, employing EEG hyperscanning to construct a dyadic EEG dataset for AIGC detection. This study employed inter-subject correlation (ISC) analysis to investigate the differences of interpersonal neural synchronization (INS). Additionally, this study proposed a novel neural network model named Squeeze-and-Excitation Depthwise Separable Convolution (SEDSC) for predicting the authenticity of real vs. AIGC. ISC analysis revealed apparent differences in INS under different modalities, valences, and animacy. Specifically, across the four frequency bands, both text and audio modalities elicited higher inter-brain synchronization under real materials than under AIGC materials. SEDSC utilized the phase locking value to assess inter-brain functional connectivity and weighted the inputs from four frequency bands before feeding them into the network for classification. This approach achieved a classification accuracy of 92.42% in distinguishing real from fake content. This study designed a new experimental paradigm and constructed a dataset, confirming that there were evident differences in INS during tasks involving real and AIGC materials. Furthermore, SEDSC successfully predicted the authenticity of the content.

Detection of AI-Generated Contents Based on Dyadic-Brain Neural Synchronization

Parkinson's disease (PD), characterized by its incurable nature, rapid progression, and severe disability, poses significant challenges to the lives of patients and their families. Given the aging population, the need for early detection of PD is increasing. In vitro diagnosis has garnered attention due to its non-invasive nature and low cost. However, existing methods present several challenges: 1) limited training data for facial expression diagnosis; 2) specialized equipment and acquisition environments required for gait diagnosis, resulting in poor generalizability; 3) the risk of misdiagnosis or missed diagnosis when relying on a single modality. To address these issues, we propose a novel multimodal in vitro diagnostic method for PD, leveraging facial expressions and behavioral gait. Our method employs a lightweight deep learning model for feature extraction and fusion, aimed at improving diagnostic accuracy and facilitating deployment on mobile devices. Furthermore, we have established the largest multimodal PD dataset in collaboration with a hospital and conducted extensive experiments to validate the effectiveness of our proposed method.

A Multimodal In Vitro Diagnostic Method for Parkinson’s Disease Combining Facial Expressions and Behavioral Gait Data

Sleep staging is the foundation of sleep analysis. Recent studies have attempted to integrate multimodal signals, such as electroencephalogram (EEG) and electrooculogram (EOG), to enhance the sensitivity of models. However, these attempts still face limitations in effectively merging multimodal signals, particularly in capturing the interplay of global and local information during sleep stages simultaneously. To address this issue, we propose a Multi-granularity Sleep Staging Network (MGSleepNet), which integrates two core modules: the Global Feature Integration module (GFI) and the Fine-grained Information Capture module (FIC). The GFI effectively captures the global features of EEG and EOG signals through multi-scale convolution, channel attention mechanisms, and spatial attention mechanisms. The FIC module obtains fine-grained interaction information between EEG and EOG by segmenting time periods and employing cross-attention mechanisms. The combination of these modules resulted in an accuracy of 83.16% and 82.46% in five-fold cross-validation across subjects on the Sleep-edf-20 and Sleep-edf-78 datasets, respectively. In addition, we also achieved better than opportunity level performance on our own data sets. Finally, the ablation studies confirmed the benefits of integrating global and fine-grained relevance paradigms to enhance sleep staging performance. The model input research indicated that MGSleepNet demonstrates good performance in sleep staging outcomes.

Downloads

Next from CogSci 2025

Perceptions of A.I.-Enhanced Bodies: Autonomy, Authenticity, and Preferences Among Young Adults

Similar lecture

Towards Fully Automated Manga Translation