Understanding physicians' noncompliance use of AI-aided diagnosis—A mixed-methods approach

Abstract

Despite the pervasiveness of artificial intelligence (AI) technologies in the healthcare industry, physicians are reluctant to follow the recommendations suggested by AI-aided diagnostic systems. We conceptualize physicians' noncompliance use of AI-aided diagnostic systems and draw on the technology threat avoidance theory (TTAT) to investigate the phenomenon of interest. Specifically, we leverage a mixed-methods approach to develop and test a comprehensive research model of physicians' noncompliance use of AI under the overarching theory of TTAT. With an exploratory qualitative study by interviewing ten physicians with experience in using AI-aided diagnostic systems, we observe that (1) physicians experience two distinct types of threats imposed by AI, namely AI threats to diagnostic process and outcome, (2) physicians' resistance to AI-aided diagnostic systems is the underlying psychological mechanism that turns their AI threat perceptions into noncompliance usage behavior, and (3) physicians' professional capital serves as an essential boundary condition in understanding the impacts of AI threats on resistance. In a confirmatory quantitative survey with 160 physicians, we find that (1) both AI threats to diagnostic process and outcome arouse physicians' psychological resistance, (2) such resistance to AI-aided diagnosis leads to noncompliance usage behavior, (3) noncompliance use of AI-aided diagnosis decreases physicians' diagnostic performance enhanced by AI, and (4) physicians' professional capital weakens the positive impact of AI threat to diagnostic process on resistance, but strengthens the positive impact of AI threat to diagnostic outcome on resistance. Our research advances the understanding of post-adoption noncompliance use of AI technology and enriches TTAT in health AI use. Our empirical findings offer practical suggestions for implementing and managing AI technology in the healthcare industry.