Projects

FureAI: A platform supporting the daily life of older adults through social conversational agents and smart behavior monitoring

Principal Investigator

Takayuki Ito

Summary

This project jointly develops FureAI ("fureai" means friendly contact in Japanese), a platform that uses AI to monitor and manage the health of older adults living alone and to build social networks among them, in order to improve their physical and mental well-being. The Canadian team researches physical-activity monitoring with AI-enabled devices, and the Japanese team develops methods to support social interaction using conversational AI. The project builds a comprehensive health-monitoring and management platform so that older adults can live comfortably at home: personal AI agents establish social networks, provide personalized feedback and dialogue, and improve physical and mental quality of life.

JST project database ↗

Collaboration with the Fukushima Institute for Research, Education and Innovation (F-REI)

Creative consensus-building algorithms for multiagent systems in complex negotiation environments

Principal Investigator

Takayuki Ito

Summary

This project realizes creative consensus-building algorithms for multiagent systems in complex negotiation environments and applies them to the real world. We develop negotiation algorithms that creatively generate new agreements even when no agreement is known in advance. In the real world it is common that, starting from a situation with no candidate agreement, people talk, collect new issues and ideas, discover an agreement that did not exist before, and reach consensus. We propose a new dynamic negotiation-space model that can represent real, uncertain situations. Using deep reinforcement learning, agents reach creative agreements that even humans do not notice. On this basis we build an agent-based consensus-building support system and apply it in practice.

KAKEN ↗

Human-Centric Autonomy from Agentic Dynamics

Principal Investigator

Rafik Hadfi

Summary

The current achievement involves surveying the two main concepts of the project (agency and autonomy) and developing algorithms that use information-theoretic metrics, precisely mutual information, to define autonomy and agency. The resulting algorithms were tested on automata that simulate systems of interacting players (in the game theoretic sense). The findings from these tests are now being prepared for submission to a journal. The second achievement is the presentation of the theoretical model of the research and its implications for the study of agency and autonomy from an enactive perspective. This theoretical model results from the surveys of the concepts in cognitive sciences, particularly in 4E cognition. The last finding was presented at the CHAIN Philosophy Workshop "Frontiers of Enactivism in East Asia" at Hokkaido University.

KAKEN ↗

AI-Mediated E-Contact for Reducing Bias and Prejudice in Intergroup Interactions

Principal Investigator

Sofia Sahab

Co-Investigators

Takayuki Ito, Jawad Haqbeen

Summary

In today's world, communication often occurs within like-minded groups, perpetuating misunderstandings, prejudice, and conflict. While traditional social psychology theories offer insights into bridging these divides, their application is limited in the digital age. Our research aims to fill this gap by integrating AI into online communication to enhance collaboration, consensus-making, and reduce social tensions. We will approach this in three phases: first, we develop a theoretical framework by translating social psychology theories into AI principles designed to reduce intergroup prejudice and anxiety while promoting teamwork; second, we integrate these principles into AI systems to foster positive interactions; and third, we conduct real-world experiments to assess the AI's effectiveness in reducing tensions and improving well-being. The research is original in its novel application of AI to enhance intergroup relations and set new standards for AI-driven social cohesion. Expected outcomes include advancing the understanding of AI for real-world issues and achieving practical impacts like enhancing global harmony, reducing prejudice and biases, and supporting peacebuilding.

Automated negotiating agents that reach agreement with humans through multiagent reinforcement learning

Principal Investigator

Shiyao Ding

Co-Investigators

Takayuki Ito, Donghui Lin

Summary

This project aims to develop automated negotiating agents that reach agreement with humans and to confirm their usefulness through real-world application. Conventional research on automated negotiating agents has limits when applied to negotiation with real people: it has focused on agent-to-agent negotiation in simulation, and does not account for the bounded rationality of real humans or the burden of many negotiation rounds. Using multiagent reinforcement learning and taking the characteristics of human bounded rationality into account, we develop agents that reach agreement within a few rounds, and verify their effectiveness in real-world trials such as online sales of electrical appliances.

Mathematical study of dialogue facilitation: modeling, and methods for judging and inferring the appropriateness of dialogue facilitation

Principal Investigator

Ryuta Arisaka

Summary

By combining knowledge about dialogue facilitation found across fields such as psychology and education with knowledge from AI and formal argumentation, this project builds a mathematical theory of (language-mediated) dialogue facilitation and, on that basis, develops methods to infer whether a dialogue is being facilitated appropriately. Three baseline goals are set: (1) identify the rational principles underlying facilitation methods whose effectiveness has been verified in various fields, formulate them as mathematical requirements, and obtain a mathematical model of dialogue facilitation on that theoretical foundation; (2) classify the knowledge on dialogue facilitation from various fields from a general viewpoint so that field-independent templates of facilitation can be generated, making it easier to transfer and adapt knowledge between fields and to understand the similarities among facilitation methods systematically; and (3) construct inference methods for evaluating the appropriateness of dialogue facilitation.

From designing adaptive facilitation threshold to summarization timeline based on IBIS: Democratizing discussion in online education conversation

Principal Investigator

Jawad Haqbeen

Summary

In this study, 1) I will develop facilitation messages database to enhance automated agent posted facilitation messages. 2) Then, I plan to propose an adaptive facilitation approach to enhance conversational agent facilitation threshold behavior. 3) Third, I will develop a discussion summarizations timeline based on Issue-based Information System (IBIS) to facilitated more informed engagement. 4) Finally, I will establish an analytical model using S-O-R framework to verify proposed Tasks (T1-T3) while conducting social experiments.

KAKEN ↗

Hyper-democracy: a large-scale consensus-building platform based on social multiagent systems

Principal Investigator

Takayuki Ito

Summary

This project realizes a platform for democracy on social networking services in which software agents and humans participate together (hyper-democracy). Agents acting on behalf of people are distributed over an SNS that serves as the infrastructure for consensus building, and they mediate the consensus-building process (a social multiagent system). While addressing problems such as flaming and fake news, the platform supports better consensus building and collective decision making.

Project site ↗ / JST CREST ↗