Appraisals of harms and injustice trigger an eerie feeling that decreases trust in artificial intelligence systems

As artificial intelligence (AI) becomes more pervasive, the concern over how users can trust artificial agents is more important than ever before. In this research, we seek to understand the trust formation between humans and artificial agents from the morality and uncanny theory perspective. We conducted three studies to carefully examine the effect of two moral foundations: perceptions of harm and perceptions of injustice, as well as reported wrongdoing on uncanniness and examine the effect of uncanniness on trust in artificial agents. In Study 1, we found perceived injustice was the primary determinant of uncanniness and uncanniness had a negative effect on trust. Studies 2 and 3 extended these findings using two different scenarios of wrongful acts involving an artificial agent. In addition to explaining the contribution of moral appraisals to the feeling of uncanny, the latter studies also uncover substantial contributions of both perceived harm and perceived injustice. The results provide a foundation for establishing trust in artificial agents and designing an AI system by instilling moral values in it.