Exploration and practice of human-machine trustworthy mechanism in XAI

Artificial Intelligence (AI) has made significant progress in various industries, but the black-box issue, potential risks, and the resulting user trust crisis have limited its further promotion and application. This paper addresses the trust issues in AI by proposing a general U-XAI (unified-trustworthy XAI) human-computer trust mechanism and technical framework, aiming to solve the eight types of trust challenges brought by the "social-technical gap" in the AI field. The framework includes four modules: trust chain governance, integrity governance, understandability governance, and acceptability governance. Combining theoretical models and technical practices, it provides a comprehensive solution for building trustworthy AI. Through the U-XAI framework proposed in this paper, the credibility of AI systems can be effectively enhanced, promoting their application and development in various social fields, and providing a practical basis and reference for the trustworthy governance of AI.