Repairing trust in AI: Attribution, anthropomorphism, and the illusion of difference.

Trust repair between humans and AI systems is an emerging area of inquiry that builds on a well-established literature in human-human (HH) interactions. This review examines how key HH findings, particularly the differential effectiveness of repair strategies following competence versus integrity violations, generalize to human-robot interaction (HRI). While some patterns hold, important divergences exist. The paper argues that better operationalizations of integrity violations and increased anthropomorphism can align HRI findings more closely with HH research. However, enduring challenges remain regarding the ethics of anthropomorphic design, the moral values AI should adopt, and how AI should extend trust to humans.