[1Win4] Poster session 1

[1Win4-95] Reevaluating the role of trust in AI society

〇Toshinari Yamashita¹, Kai Terakado¹ (1. GMO BRAND SECURITY Inc.)

Keywords:Trusted Role, Institutional and Human Trust, Social Complexity

This study considers how to manage the social complexity and risk caused by the spread of generative AI (e.g. ChatGPT) while referring to the "Trusted Role" model in PKI (Public Key Infrastructure) certification authorities. Using Luhmann's theory of social complexity reduction and Beck's risk society theory as a theoretical framework, we propose a mechanism that enables bridging institutional trust (performance and audit) and human trust (ethics and responsibility) by applying the concept of "role assignment" shown in traditional role theory to AI.
In particular, we refer to a case study of the fusion of cryptographic technology and human auditing in PKI, and demonstrate the significance of applying this to "role design" of generative AI. As a result, we suggest that an approach that defines clear roles and establishes a system of responsibility and auditing within those roles, rather than using AI unlimitedly, may suppress the amplification of social complexity and lead to the safe use of AI.