表題番号：2024C-182
日付：2025/04/02
研究課題福祉機器開発時の安全情報共有における生成AI活用の可能性に関する研究
| 研究者所属（当時） | 資格 | 氏名 | |
|---|---|---|---|
| （代表者） | 理工学術院 大学院基幹理工学研究科 | 教授 | 本間 敬子 |
- 研究成果概要
- Developing assistive devices requires collaboration among stakeholders with diverse backgrounds, including manufacturers, users, and medical professionals. A major challenge in this process is ensuring effective communication and information sharing among individuals with varying expertise. It is expected that utilizing generative AI to, so to speak, 'translate' words among people with different backgrounds can help address this issue. This study aims to develop a system that supports the safe and efficient sharing of safety-related information regarding assistive devices.One of the significant concerns in using generative AI for dialogue generation is hallucination—a phenomenon where AI generates plausible but factually inaccurate content. This can pose serious risks, especially when handling safety-related information. Retrieval-Augmented Generation (RAG) is a promising approach to mitigating this issue, as it enhances reliability by referencing external sources. By leveraging RAG, this study seeks to develop a robust information-sharing system that minimizes misinformation risks.To achieve this goal, the study involves selecting appropriate tools and establishing a computational environment optimized for safe data handling. Additionally, a literature review has been conducted to identify credible sources that can be integrated into the system. Ensuring the accuracy and trustworthiness of the information shared is critical, as errors in safety-related content can lead to severe consequences. By incorporating RAG-based mechanisms, the system can provide stakeholders with verifiable and well-supported responses.Moving forward, the study will focus on refining the system’s implementation and assessing its effectiveness in real-world scenarios. The ultimate goal is to create a reliable framework that supports informed decision-making among stakeholders involved in assistive device development. By continuously improving the methodology and integrating trustworthy data sources, this research aims to contribute to safer and more efficient collaboration in the field.