Ken Satoh

Ken Satoh graduated from the Department of Information Science, Faculty of Science, University of Tokyo (March 1981) and joined Fujitsu Laboratories in April of the same year. He obtained a Doctor of Science (Ph.D.) from the University of Tokyo in January 1993. In July 1995, he became an associate professor at Hokkaido University and was appointed professor at the National Institute of Informatics (NII) in April 2001. In November 2023, he was named Director of the Legal Informatics Centre at the Joint Research Support Centre for Data Science, Research Organisation of Information and Systems (ROIS-DS), before becoming full-time Director in April 2024.

His research focuses on the foundations of legal informatics (AI and law) and logic-based artificial intelligence. From 2006 to 2009, he studied law at the University of Tokyo, Graduate School of Law and Politics, to explore the application of artificial intelligence to legal studies. He obtained his bar qualification in 2017.

The project

Title: Exploration France

"With the development and dissemination of AI technologies, ensuring that AI systems comply with legal and ethical principles has become a major challenge. Concerns about the undesirable effects of these systems—through their actions and their use of personal data—have increased demand for reliable AI, making this issue a priority both in public opinion and on the policy agenda.

Technical solutions are therefore needed, and it is widely recognized that mechanisms addressing these concerns should be integrated into the core architecture of AI agents. In this context, I have led a trilateral collaborative research project (Japan-France-Germany) on AI compliance, the RECOMP project (Realtime Compliance Mechanism for AI, 2021–2024), together with researchers from Sorbonne University and the Fraunhofer Institute. The system developed has proven useful, but its implementation, based on a logical framework, reveals discrepancies between standards formulated in natural language and their representation as logical formulas.

The aim of this proposal is to develop a translation of these standards from natural language into logical formulas using large language models. My research group is already working on processing legal documents with these models, and the expected results will contribute to the development of a reliable translation method applicable to AI systems."

Hosting Institution: Sorbonne University

Selective Bibliography

• May Myo Zin, Satoh, K., Borges, G., "Leveraging LLM for Identification and Extraction of Normative Statements", Proc. of JURIX 2024, pp. 215 - 225 (2024).
• Fungwacharakorn, W., Takeda, H., and \me, "Using WikiData for Handling Legal Rule Exceptions: Proof of Concept", New Frontiers in Artificial Intelligence, LNAI 14644, pp.85-99 (2024).
• May Myo Zin, Nguyen, H.T., \me, Nishino, F., "Addressing Annotated Data Scarcity in Legal Information Extraction", Proc. of JSAI-IsAI 2024, LNCS 14741, pp.77-92 (2024).
• Zin, M.M., Nguyen, H.T., \me, K., Sugawara, S., Nishino, F., "Information Extraction from Lengthy Legal Contracts: Leveraging Query-Based Summarization and GPT-3.5", Proc. of JURIX 2023, pp. 177 - 186 (2023).
• Zin, M.M., Nguyen, H.T., \me, K., Sugawara, S., Nishino, F., "Improving Translation of Case Descriptions into Logical Fact Formulas using LegalCaseNER", Proc. of ICAIL '23, pp. 462 - 466 (2023).