Research on the Influencing Factors of User Trust Based on Artificial Intelligence Self Diagnosis System

Abstract

Artificial intelligence technology has gradually become an important auxiliary means for people to obtain medical consultation. However, many researches reflect users' doubts about the feedback results of AI system. Therefore, this paper focuses on the design of the user-friendly intelligent self diagnosis system, using interdisciplinary research methods. Firstly, based on the psychological questionnaire and interview, this paper studies the public's cognition of the intelligent self diagnosis system, and discusses how to effectively improve the user's recognition and trust of the intelligent self diagnosis system. It is found that the credibility of the intelligent self diagnosis system can be improved by further improving the transparency of the system. For example, users can be provided with four different types of information, including system reasoning, system reliability, information source and personalized information. On the basis of interviews, 48 volunteers were invited to carry out a comparative experiment in groups to evaluate the impact of medical interpretation of intelligent self diagnosis system on patients' perception and trust under different transparency and accuracy models. We found that the improvement of system reliability information (i.e. accuracy model and confidence score) composed of accuracy model and confidence score strengthens patients' cognition of AI system and improves the user trust of the system; while the detailed display of system reasoning and logic can improve the comprehensibility of medical advice output of the system, but does not substantially improve the user trust. At the end of this paper, some suggestions are put forward to design an interpretable and reliable intelligent self diagnosis system.