Consensus Informatics Field, Division of Social Information Network
Department of Social Informatics, Graduate School of Informatics, Kyoto University
Ito Laboratory
AI-powered Collective Intelligence
With the usage rate of SNS and smart devices increasing dramatically, a fundamental modification of the interaction among us is being presented. The social system we daily encounter is the collective intelligence constructed with human wisdom. However, almost all of the social systems are classical, created in the era when SNS and smart devices had not come out. Meanwhile, groups of fish and insects are evolving their collective intelligence at a dramatic speed due to the need for evolutionary survival. We need to implement a new social system to promote the collective intelligence of humankind.
Multi-agent systems provide methodologies and concepts for implementing new social systems and promoting human collective intelligence with intelligent information technology. In the field of multi-agent systems, the main focus is on exploring the nature of social intelligence, exploring the possibilities of new social systems and implementing them in society. Specifically, interdisciplinary research is being developed in the fields of distributed artificial intelligence, simulation, robotics and game theory.
Our research focuses on the theory, modelling, simulation and social implementation of multi-agent systems. In particular, we conduct advanced research on consensus-making support, computational mechanism design, automatic negotiation agents and social simulation.
We pursue both theoretical research that aims for high ideals and business and commercialisation that faces up to harsh realities.
Members
In the Laboratory of Prof. Takayuki Ito, teachers and students work together to improve 'humanity', 'creativity', 'intelligence', 'physical strength' and 'internationality'.
Professors & Staff
Professor
Takayuki Ito
Associate Professor
Rafik Hadfi
Lecturer
Ryuta Arisaka
Assitant Professor
Shiyao Ding
Assitant Professor
Sofia Sahab
Assitant Professor
Jawad Haqbeen
Secretary
Rika Ikebe
Students
Yu Kimura D3
DONG Yihan D2
Keijiro Asama M2
Tatsuki Ikeda M2
Kazuki Matsumoto M2
Kazuhito Mori M2
Uichi Suzuki
M2
LIU Yuchen M2
Hideki Kitagawa
M2
M2
Lita Tang M
Kengo Hayashi M
Hokuto Wakita M2
Takafumi Shimizu
M1
M1
Kento Ueda
M1
M1
Yuta Konishi
M1
M1
Keiichiro Nii
M1
M1
Kengo Hayashi
M1
M1
Soichiro Tanihata
B4
B4
Takumi Komukai
B4
B4
Mikito Kanaomori
B4
B4
WANG Siyuan RS
ZHOU Zibo RS
Yuki Ito M2
(Ma laboratory)
Yuto Sakai M2
(Ma laboratory)
Takuto Nabeoka M2
(Ma laboratory)
People Who Collaborate with Us
Tokuro Matsuo, AIIT, CREST
Shun Shiramatsu, NITech, CREST
Susumu Ohnuma, Hokkaido University, CREST
Shun Okuhara, Mie University
Wen Gu, JAIST
Donghui Lin, Okayama University
Main Research Theme
Agent-based Crowd Decision Support
Supporting the crowded decision-making by using agent technology. The systems called "Collagree" and "Dagree" have been implemented and used in the real world. Developing upon this, the current CREST project is researching new forms of democratic platforms.
Automated Negotiating Agents
This research is concerned with agents that could negotiate automatically. The main activities are exploring the negotiation theory based on the fundamental issue of what negotiation is, designing negotiation models and implementing negotiation agents based on the models.
Computational Mechanism Design
We design theories of mechanisms in which truth declarations are best. We design social institutions (mechanisms) using game theory and microeconomics, but in particular, we design new mechanisms from an informatics perspective or clarify and resolve issues in classical theories.
Industrial Application with AI technologies
It conducts a wide range of industrial cooperation activities and research with developing software to solve problems in industry in a sophisticated way. Recently, it has been working in particular with deep learning.
Argumentation Theory
This is a mathematical-theoretical study of logical arguments between agents. The research is not limited to the existing logic of argumentation, but also includes the acceptability of arguments due to their influence, with a view to their application to actual argumentation.
Multiagent Simulation
Implementation simulation of large-scale social systems by using multi-agents and a set of micro-level activity models. Furthermore, we also aim at a more unified and comprehensive simulation by linking social and physical systems with a multi-layered structure.
Multiagent DRL for Cooperation
Multiple entities are inherently superior when cooperation can be achieved appropriately. To study the emergence of such cooperation, we are currently working on multi-agent deep reinforcement learning.
Decision Making, Group Decision Making and Consensus
The project will study human decision-making, group decision-making and consensus-reaching, as well as research and development on how to support such decision-making. Particular attention will be paid to social networks and crowds, and also to the challenge of how to estimate and extract individual people's preferences and opinions.
Check Out The Video
This is the promotional video for D-agree, which is a large-scale discussion support system based on automated facilitation agents. It could extract valuable opinions, structure discussions and support decision-making.
This is the video introducing our laboratory.
Research Projects
Research Theme：Hyperdemocracy: a large-scale consensus-building platform based on social multi-agents.
Research Overview： This research will realise a platform for democracy (hyper-democracy) in SNSs, where software agents and humans participate together. Here, human surrogate agents are distributed in the SNS as a foundation for consensus building and mediate the consensus building process (social multi-agent system). It then supports better consensus-reaching and collective decision-making while resolving issues such as flame wars and fake news.
For details, please check here
Research Theme： Multi-agent creative consensus-reaching algorithms in complex negotiation environments.
Research Overview： In this research, a multi-agent creative consensus-reaching algorithm in complex negotiation environments is implemented and applied in real-world applications. Here, we realise a negotiation algorithm that creatively generates new agreement proposals even in situations where the agreement proposal is not clear in advance. For example, the examples that people discuss with each other to gather new issues or ideas, and further find and reach an agreement that did not exist initially in a situation where was no draft agreement at first, often happen in the real world. In this study, we propose a new dynamic negotiation space model that can represent real-life uncertain situations. The agent uses deep reinforcement learning to achieve creative agreements that even humans are unaware of. Using the above, we realise an agent-based consensus support system and apply it to the real world.
Student Recruitment
Selection Schedule for Research Students
Deadlines for application to AAO
for April enrollment: at the end of September.
for October enrollment: at the end of March.
Notification for an interview for shortlisted candidates
for April enrollment: by mid-October.
for October enrollment: by mid-April.
Interviews
for April enrollment: in late October.
for October enrollment: in late April.
Notification of acceptance
for April enrollment: in early November.
for October enrollment: in early May.
Each application is evaluated on CV, communication ability, and research proposal.
Contact.
Please contact us with the information followed:
Yoshida Honmachi, Sakyo-ku, Kyoto 606-8501 Japan. Division of Social Information Network, Department of Social Informatics, Graduate School of Informatics, Kyoto University
+81 075-753-4821 ito [at] i.kyoto-u.ac.jp
If you are interested in our laboratory and want to talk in English, then please contact with the following email address:
contact[at]agent.soc.i.kyoto-u.ac.jp