CREST

Japan Science and Technology Agency Strategic Basic Research Programs
Strategic Basic Research Programs

[Symbiotic AI Interdisciplinary System] Year Started : 2025

Tetsunari Inamura

Creating a Proactive Society through Behavior Transformation in Co-living with Self-Mirroring Twins

Grant No.:JPMJCR2561

Research Director :Tetsunari Inamura
Photo:Tetsunari Inamura

Professor
Brain Science Institute
Tamagawa University

Collaborator
Minoru AsadaSpecially-Appointed ProfessorOpen and Transdisciplinary Research InitiativesThe University of Osaka
Makoto KurehaAssociate ProfessorFaculty of Global and Science StudiesYamaguchi University
Shingo ShimodaDesignated ProfessorGraduate School of MedicineNagoya University
Takato HoriiAssociate ProfessorGraduate School of Engineering ScienceThe University of Osaka
Outline

We will establish a methodology for applying Self Mirroring Twins (SeMT) —a twin-like agent whose self-other boundary can be continuously tuned— to support lifestyle improvement and rehabilitation. We will demonstrate the effectiveness of technologies that foster self-directed behavior change. The project comprises five interlinked research components: platform development, rehabilitation-assist applications, computational models of behavior change, evaluation of brain activity and biosignals, and the formulation of ELSI guidelines. Working in concert, these components will implement a future social model in which coexisting with AI makes personal growth easier.

Kazutoshi Sasahara

Trust-Oriented Social AI Design for Overcoming an Uncertain Society

Grant No.:JPMJCR2562

Research Director :Kazutoshi Sasahara
Photo:Kazutoshi Sasahara

Professor
School of Environment and Society
Institute of Science Tokyo

Collaborator
Tasuku IgarashiProfessorGraduate School of Education and Human DevelopmentNagoya University
Isao EchizenProfessorNational Institute of InformaticsResearch Organization of Information and Systems
Outline

In today’s society, where the spread of fake news undermines social trust, this study aims to establish a design theory for trust-oriented social AI, toward realizing a symbiotic society in which humans and AI collaborate based on trust. By introducing a mechanism that enables AI to estimate and adjust trust readiness, it promotes the co-creative construction of trust. Through interdisciplinary research in computational social science, cognitive security, and social and personality psychology, this study will develop a new information infrastructure to strengthen societal resilience against fake news.

Kazunori Terada

Resolving Real-World Social Dilemmas through Reliable and Safe Theory of Mind in AI

Grant No.:JPMJCR2563

Research Director :Kazunori Terada
Photo:Kazunori Terada

Professor
Faculty of Engineering
Gifu University

Collaborator
Shogo OkadaProfessorGraduate School of Advanced Science and TechnologyJapan Advanced Institute of Science and Technology
Hirokazu KumazakiProfessorGraduate School of Biomedical SciencesNagasaki University
Yukiko NakanoProfessorFaculty of Science and TechnologySeikei University
Outline

This research aims to develop AI that facilitates human win-win decision-making and resolves social dilemmas ubiquitous in real society. The project will achieve: accurate inference of individuals’ multidimensional, complex, and dynamic cognitive-affective states; multi-objective optimization generating win-win solutions with high satisfaction levels; and generation of behavior change signals that ensure psychiatric safety while preserving individual autonomy. Through this approach, we will establish a formal framework that guarantees the reliability and safety of theory of mind in AI. The approach will be validated through game-theoretic scenarios, customer support applications, and high-stakes shared medical decision-making tasks.

Itsuki Noda

Community System Design by MASS with Mesoscopic Hierarchical Modeling

Grant No.:JPMJCR2564

Research Director :Itsuki Noda
Photo:Itsuki Noda

Professor
Graduate School of Information Science and Technology
Hokkaido University

Collaborator
Munenari InoguchiProfessorCollege of Policy ScienceRitsumeikan University
Ryo KanamoriDesignated professorInstitute of Innovation for Future SocietyNagoya University
Nobutsugu KanzakiProfessorFaculty of Global Liberal StudiesNanzan University
Yuko SakuraiProfessorGraduate School of EngineeringNagoya Institute of Technology
Hiromitsu HattoriProfessorCollege of Information Science and EngineeringRitsumeikan University
Outline

We aim to establish MASS technologies incorporating mesoscopic hierarchical modeling, game-theoretic analysis, and simplifying techniques for human activities in local communities. Building on these methods, we will develop an engineering framework that facilitates consensus-building and continuous improvement in the design of community services and policy-making, including contexts where people coexist with AI. As concrete case studies, we will address transportation, disaster prevention, and urban development, and will construct methods of computational social engineering not only as technical mechanisms but also as new tools in the humanities and social sciences for envisioning future societies.

Daisuke Bekki

The Nexus of Human-LLM Interaction: A Multidisciplinary Linguistic Approach to Building a Collaborative Feedback Loop

Grant No.:JPMJCR2565

Research Director :Daisuke Bekki
Photo:Daisuke Bekki

Professor
Faculty of Core Science
Ochanomizu University

Collaborator
Yohei OsekiAssociate ProfessorGraduate School of Arts and SciencesThe University of Tokyo
Daisuke KawaharaProfessorFaculty of Science and EngineeringWaseda University
Takuya MatsuzakiProfessorFaculty of Science Division ITokyo University of Science
Hitomi YanakaAssociate ProfessorGraduate School of Information Science and TechnologyThe University of Tokyo
Outline

Despite their capabilities, large language models (LLMs) face significant challenges in validation and improvement when applied to real-world tasks. This research proposes a solution using linguistic pipelines as a mediating framework. First, we will investigate the performance limits of both LLMs and linguistic pipelines in natural language understanding. Second, we will establish a feedback loop where linguistic pipelines, which are manageable by linguists, can validate the reasoning of LLMs and provide a reward model for improvement. This framework creates a collaborative workflow where linguistic experts work alongside engineers throughout the development and operational lifecycle of LLMs, ensuring a principled and verifiable approach to their deployment.

Taro Maeda

Establishing Nonverbal Communication Channels to Promote Human AI Coexistence and Collaboration

Grant No.:JPMJCR2566

Research Director :Taro Maeda
Photo:Taro Maeda

Professor
The Graduate School of Information Science and Technology
The University of Osaka

Collaborator
Hiroyuki IizukaAssociate ProfessorCenter for Human Nature, Artificial Intelligence, and NeuroscienceHokkaido University
Norihiko SagaProfessorSchool of EngineeringKwansei Gakuin University
Outline

The objective of this research is to address the challenges of achieving mutual understanding between humans and AI by leveraging non-verbal communication channels, particularly body language. Through proprietary technology that learns and externalizes the wearer’s bodily responses, an auxiliary non-verbal network is being built to support collaboration between humans and AI, between AI systems, and between multiple humans and multiple AI systems. By enabling support for mutual understanding that does not rely solely on verbal communication, we aim to advance the realization of a society in which humans and AI coexist and collaborate.

Page Top