Fairness, Accountability, Transparency, Explainability: A Qualitative Approach towards Trustworthy AI in Autonomous Vehicles

Abstract

A successful market launch of autonomous vehicles (AV) is only possible if users trust the AV and thus the artificial intelligence (AI) powering the vehicle. To conceptualize trust in AI, researchers recently started using so-called FATE characteristics (Fairness, Accountability, Transparency, Explainability). Until now, the FATE characteristics have not been contextualized by AV specific FATE attributes. This work aims to answer how to establish trust with the FATE characteristics in AVs by conducting a content analysis of 33 AV provider websites and 5 expert interviews. The findings suggest that in the context of AVs, the C-FATS characteristics (Certifiability, Fairness, Accountability, Transparency, Safety) better conceptualize trust. Applying the results of the analysis, a framework for TAI in the context of AVs encompassing 91 C-FATS attributes is developed. In addition, differences between providers and experts in the conceptualization of TAI in AVs are highlighted and interdependencies that need to be considered in establishing TAI are identified. Since the interdependencies between trust-building attributes challenge the distinction between “trust in technology” and “trust in organizations”, researchers are tasked to generalize extended trust concepts in the context of AI systems to include transfer of trust.