A Critical Analysis of the Trust Human Agents Have in Computational and Embodied AI

Abstract

The notion of trust has evolved to become one of the many nebulous buzzwords centering around AI. This paper aims at showing that, with regards to human influenceability through AI, trust in AI is to be seen as problematic. Based on the notion of socio-technical epistemic systems, I will argue that the trust human agents have in AI is strongly related to what could be understood as algorithmic authority. The second part of this paper will then translate the elaborated line of argument to the field of social robotics.