Abstract: With the development of technology, artificial general intelligence has begun to take shape, ushering in a new era for human-machine interaction and relationships. The trust between humans and artificial intelligence (AI) are on the brink of a transformative shift from unidirectional trust, where people trust AI, to a state of mutual trust between humans and AI. This study, based on a review of the interpersonal trust model in social psychology and the human-machine trust model in engineering psychology, proposes a dynamic mutual trust model for human-AI relationships from the perspective of interpersonal trust. The model regards humans and AI as equal contributors to trust-building, highlighting the “mutual trust” in the relational dimension and the “dynamics” in the temporal dimension of trust between humans and AI. It constructs a fundamental framework for dynamic mutual trust between humans and AI, incorporating influencing factors, result feedback, and behavior adjustment as essential components. This model pioneers the inclusion of AI’s trust towards humans and the dynamic interactive process of mutual trust, offering a new theoretical perspective for the study of trust between humans and AI. Future research should focus on understanding the establishment and maintenance of trust from AI towards humans, developing quantitative models for human-AI trust, and exploring mutual trust dynamics within multi-agent interactions.

Key words: trust, human-machine mutual trust, trust calibration, human-machine relationship, human-AI