Prof. HU Chuan-Peng：Self-Bias and Its Boundaries: a Computational Perspective
In the past decade, psychologists have accumulated substantial evidence for self-bias—the phenomenon that people process self-related information faster and more accurately. However, computational models of this effect are still lacking. Within the framework of evidence accumulation models, we studied self-bias using the self-matching task and found that it may result from more efficient information uptake. To explore the boundaries of self-bias, we developed a novel computational framework that incorporates both the behavioral data and the experimental designs under which these data were produced. Our initial results suggest that self-bias effects, as measured by the self-matching task, vary across the experimental design space. This computational framework provides new insights into the cognitive mechanisms underlying self-bias.
Dr. DING Yi：Rethinking Self-Evaluation: a Socially Grounded Framework
While self-evaluation has long been conceptualized as an internal cognitive process, recent evidence suggests that it is dynamically regulated by social relationships. Previous studies have shown that social feedback—such as acceptance and rejection—can influence how individuals evaluate themselves. However, the neural mechanisms linking social relationships to self-evaluation remain unclear. To address this gap, we propose a socially grounded framework of self-evaluation that captures its dynamic and adaptive nature within social contexts. Our results provide empirical support for a dynamic model in which self-evaluation is socially embedded and adaptively regulated. This self-evaluation framework offers a new direction for understanding the interplay between mind and society.
（休憩 20分）
Dr. LIU Chunlin：Conversing with AI inside the Scanner: a Novel fMRI Paradigm for Studying Real-Time Human–AI Dialogue
Recent advances in large language models (LLMs) have enabled real-time human–AI interaction, creating new opportunities for studying communication in ecologically valid yet controlled settings. This study introduces a novel LLM–fMRI paradigm allowing participants to engage in free, turn-by-turn English conversation with ChatGPT inside the MRI scanner. Fourteen Japanese learners of English participated under three feedback conditions: no feedback, explicit correction, and implicit recast. Our results revealed activation in language-related and socio-emotional regions, indicating that AI-delivered feedback dynamically modulates both linguistic and emotional processing. This methodological innovation demonstrates that real-time human–AI conversation can be successfully implemented in fMRI, bridging ecological validity and experimental precision. The paradigm opens new avenues for neurocognitive research on human–AI communication, adaptive feedback processing, and the future integration of AI-based dialogue systems into cognitive neuroscience and language education.
計算科学と心理学理論の急速な発展により、人間の心と社会現象を多角的に探究するための新たな方法が広がっています。本シンポジウムでは、理論・実験・計算の視点を融合しながら、心と社会の相互作用をより深く理解するための新しい方法論的アプローチを取り扱いました。
KeynoteとしてHu Chuan-Peng教授にご登壇いただき、新しい計算モデルを用いたデータ分析の方法についてご紹介いただき、また新しい実験パラダイム、データ解析手法、理論的枠組みなど、幅広い研究アプローチについて議論しました。
心と社会の理解：理論的・実験的・計算的イノベーション
日時：10月16日（木）15:30〜18:00
場所：人間脳科学分野 SA棟3階
形式：ハイブリッド開催(現地＋Zoom）
プログラム
- Prof. HU Chuan-Peng：Self-Bias and Its Boundaries: a Computational Perspective
In the past decade, psychologists have accumulated substantial evidence for self-bias—the phenomenon that people process self-related information faster and more accurately. However, computational models of this effect are still lacking. Within the framework of evidence accumulation models, we studied self-bias using the self-matching task and found that it may result from more efficient information uptake. To explore the boundaries of self-bias, we developed a novel computational framework that incorporates both the behavioral data and the experimental designs under which these data were produced. Our initial results suggest that self-bias effects, as measured by the self-matching task, vary across the experimental design space. This computational framework provides new insights into the cognitive mechanisms underlying self-bias.- Dr. DING Yi：Rethinking Self-Evaluation: a Socially Grounded Framework
While self-evaluation has long been conceptualized as an internal cognitive process, recent evidence suggests that it is dynamically regulated by social relationships. Previous studies have shown that social feedback—such as acceptance and rejection—can influence how individuals evaluate themselves. However, the neural mechanisms linking social relationships to self-evaluation remain unclear. To address this gap, we propose a socially grounded framework of self-evaluation that captures its dynamic and adaptive nature within social contexts. Our results provide empirical support for a dynamic model in which self-evaluation is socially embedded and adaptively regulated. This self-evaluation framework offers a new direction for understanding the interplay between mind and society.（休憩 20分）
- Dr. LIU Chunlin：Conversing with AI inside the Scanner: a Novel fMRI Paradigm for Studying Real-Time Human–AI Dialogue
Recent advances in large language models (LLMs) have enabled real-time human–AI interaction, creating new opportunities for studying communication in ecologically valid yet controlled settings. This study introduces a novel LLM–fMRI paradigm allowing participants to engage in free, turn-by-turn English conversation with ChatGPT inside the MRI scanner. Fourteen Japanese learners of English participated under three feedback conditions: no feedback, explicit correction, and implicit recast. Our results revealed activation in language-related and socio-emotional regions, indicating that AI-delivered feedback dynamically modulates both linguistic and emotional processing. This methodological innovation demonstrates that real-time human–AI conversation can be successfully implemented in fMRI, bridging ecological validity and experimental precision. The paradigm opens new avenues for neurocognitive research on human–AI communication, adaptive feedback processing, and the future integration of AI-based dialogue systems into cognitive neuroscience and language education.