The 8th Meeting for Theories Group was held on June 14, 2026. A total of nine members attended the meeting: Toru Ohira, Kenta Ohira, Reiji Suzuki, Tetsuki Tamura, Shinhaeng Kim, Wanwan Zheng, Yasuko Nakamura, Hideki Ohira, and Akane Hayanagi. The meeting focused primarily on reports on the Theories Group’s research progress, confirming collaboration among the project groups, and coordination of future activities.
At the beginning of the meeting, Nakamura summarized the Theories Group’s activities since April. She particularly emphasized the importance of integrating the research outcomes of the various groups into a shared output. She also provided updates on the preparations for the General Research Meeting scheduled for August 25 and the international symposium planned for May 2027, tentatively titled “Gender and Theoretical Collaboration,” and confirmed the schedule for upcoming activities.
Research Reports
🌟 Toru Ohira and Kenta Ohira (Applied Mathematics and Delay Theory)
Toru Ohira and Kenta Ohira presented research findings on exact solutions for delay differential equations and their applications. They derived exact solutions for delay theoretical equations and introduced new developments in applied mathematics, as well as introducing applied research in origami engineering and topology. They demonstrated the potential for interdisciplinary research combining mathematical analysis with engineering and design, such as the development of foldable structures for use during disasters and non-rubber tires.
🌟 Reiji Suzuki (Artificial Life and Evolutionary Simulation)
Suzuki reported on research that incorporated environmental factors into evolutionary models of language-using agents. By simulating the competition and evolutionary processes among agents characterized by language and introducing environmental agents (biomes), he confirmed that environmental diversity increases the complexity of evolution and contributes to the maintenance species diversity.
He further reported on attempts at evolutionary optimization of large language models (LLMs), conducting experiments in which weight parameters were evolved using a genetic algorithm. A trade-off was discovered between knowledge tasks and theory of mind tasks. In addition, changes in the conceptual structure of AI were analyzed through the visualization of internal representations via sparse encoding.
🌟 Tetsuki Tamura (Political Philosophy and Democratic Theory)
Tamura presented a study on “The Possibility of a Post-Anthropocentric Democracy.” After reexamining the relationship between capitalism and democracy and offering a theoretical overview of deliberative and participatory democracy, he explored the legitimacy of decision-making that includes non-human actors (objects and the environment). He defined a “decision-making body” as “a legitimate decision accompanied by collective binding force,” and argued that the absence of binding force indicates the incompleteness of political decisions.
🌟Shinhaeng Kim (Economic Sociology and ANT Analysis)
Kim presented “An ANT Analysis of the Formation Process and Assetization of NFT Marketplaces.” Drawing on the development and operation of NFT trading platforms as a case study, he analyzed the relationships between human and non-human actors (technology and devices) from an ANT perspective. The research traced the step-by-step development process from proof of concept to C2C implementation and service expansion, revealing the interventions of non-human actors such as AI-based NFT screening (to prevent copyright infringement). He also proposed a new concept, “Exploratory Assetization,” and analyzed the interaction between operators’ market perceptions and technological devices.
🌟Wanwan Zheng (Psycholinguistics and AI Language Model Comparison)
Zheng reported on two studies.
(1) Word Difficulty and Psychological Factors
An analysis of word difficulty was conducted using both experiential factors (frequency and morphology) and psychological factors (familiarity and abstractness), combining random forest, logistic regression, and structural equation modeling. The results confirmed that these two factors influence each other. Indirect effects mediated by psychological factors were also observed, and experimental verification using reaction times is planned for future research.
(2) Comparison of Lexical Representations between Humans and LLMs
A comparison of familiarity ratings between humans (native speakers and second-language learners) and LLMs (ChatGPT, Gemini, and Claude) revealed that LLMs tend to rely heavily on concreteness and tend not to reflect embodiment. Humans, by contrast, prioritize sensory and motor elements, while second-language learners showed a tendency to rely on visual complexity.
(Authorship: Wanwan Zheng, Humanity Center for Anthropocenic Actors and Agency, Graduate School of Humanities, Nagoya University)