Abstract:

The widespread application of artificial intelligence （AI） has engendered ambivalent trust attitudes among users. On the one hand， the optimistic expectations of technological progress and the anthropomorphic characteristics of artificial intelligence predispose users to trust AI systems； on the other hand， uncertainties arising from inadequate temporal verification and latent risks simultaneously cultivate user skepticism toward AI technologies. However， existing research has predominantly focused on assessing whether artificial intelligence possesses trustworthiness qualifications and evaluating its credibility， while paying comparatively less attention to trustors’ （users’） intrinsic trust attitudes and their practical ramifications. In this regard， introducing “discretionary power” as a key analytical lens for examining the structure of trust relationships facilitates shifting research focus toward users’ entrustment behaviors， thereby offering a more effective response to the dual predicament of trust. Accordingly， it is proposed that AI system design incorporate trust-level-based access control strategies to continuously assess users’ trust levels and provide personalized services that respect and respond to users' varying degrees of trust in artificial intelligence.

Key words: artificial intelligence, trust, discretionary power, access control