Human AI Collaboration for Trust Management

Abstract

Trust is one of the principles that human-AI teams must attain for the fulfillment of their mission. Ex-plainable AI and the principle of computational re-liabilism provide AI-intrinsic solutions for trust management. When human-AI collaboration breaks down, human-AI teams turn to common sense and intuition to recover trust. In addition, research on earlier innovations has shown that institutional and organizational mechanisms such as citizen adviso-ry boards and standardization promote trust. This paper sketches a framework for deployable and actionable trust management mechanisms. To that end, it will: (1) Identify three dimensions of trust. (2) Examine the role of heterogeneous stakehold-ers in human-AI systems. (3) Address the links among interpersonal trust, institutional trust, and trust in algorithms. (4) Suggest that stakeholder heterogeneity is a multi-level and multi-faceted imperative for establishing trust in human-AI teams.