The Impact of AI Trust Violation on Trustworthiness: An Empirical Study Based on AI Chatbots

Two 3X2X2 experiments were conducted, three types of trust violations (functional trust violations, helpful trust violations, reliable trust violations)  × responsible attribution (internal attribution, external attribution)/technical attribution (temporary problems, permanent problems) × trustworthiness (affective trust, cognitive trust). The results found that trust violations were negatively affect consumer affective trust and cognitive in AI chatbots. Responsible and technical attribution moderated the relationship between trust violations and consumer affective trust in AI chatbots. An interesting finding is that neither responsibility nor technical attribution moderates trust violation and cognitive trust. This suggests that the mechanisms by which trust violation affects affective and cognitive trust are different. This study categorized trust violation into three sub-dimensions and proposed the way of technical attribution, which have important theoretical contribution as well as practical significance to the study of trust violation and trust repair in AI chatbots.