关键词: 可信AI实践, 影响因素, 驱动模型, 伦理治理

Abstract: With the promoted status of artificial intelligence (AI) strategic technologies and their enriched application scenarios, the ethical risks of AI have drawn wide attention from all social circles, who have gradually agreed to develop trustworthy AI (TAI) with enterprises acting as the principals of the practice. Based on Proceduralised Grounded Theory, the study investigates the deep-seated factors influencing enterprises’ trustworthy AI practice through in-depth interviews and supplementary secondary data. The results show that four main categories of TAI practice-willingness, TAI practice mission, TAI practice institutional environment, and TAI practice capability, have statistically significant effects on enterprises’ trustworthy AI practice; TAI practice willingness and practical capability are internal driving factors; the institutional environment of TAI practice serves as an indirect factor to regulate the relationship between an enterprise’s willingness to practice trustworthy AI and its actual behavior; the mission of trustworthy AI practice serves as an indirect factor to regulate the relationship between an enterprise’s ability to practice trustworthy AI and its actual behavior; and the constituent factors under each category have different ways of influence on variables with different functions. The study shows that the “Two Factor Driven model” (TFD model) is of critical importance for building a multi-agent and full-process TAI practice ecosystem.

Key words: trustworthy AI practice, influencing factors, driven model, ethical governance