Are you willing to forgive AI? Service recovery from medical AI service failure

Purpose As an emerging technology, medical artificial intelligence (AI) plays an important role in the healthcare system. However, the service failure of medical AI causes severe violations to user trust. Different from other services that do not involve vital health, customers' trust toward the service of medical AI are difficult to repair after service failure. This study explores the links among different types of attributions (external and internal), service recovery strategies (firm, customer, and co-creation), and service recovery outcomes (trust). Design/methodology/approach Empirical analysis was carried out using data ( N = 338) collected from a 2 × 3 scenario-based experiment. The scenario-based experiment has three stages: service delivery, service failure, and service recovery. The attribution of service failure was divided into two parts (customer vs. firm), while the recovery of service failure was divided into three parts (customer vs. firm vs. co-creation), making the design full factorial. Findings The results show that (1) internal attribution of the service failure can easily repair both affective-based trust (AFTR) and cognitive-based trust (CGTR), (2) co-creation recovery has a greater positive effect on AFTR while firm recovery is more effective on cognitive-based trust, (3) a series of interesting conclusions are found in the interaction between customers' attribution and service recovery strategy. Originality/value The authors' findings are of great significance to the strategy of service recovery after service failure in the medical AI system. According to the attribution type of service failure, medical organizations can choose a strategy to more accurately improve service recovery effect.