What we talk about when we talk about trust: Theory of trust for AI in healthcare

TLDR

Artificial intelligence is driving medical innovation, and trust is widely regarded as essential for its successful adoption. The authors propose developing a rigorously grounded conceptual framework of trust in health‑related AI to guide design and acceptance. The study finds that existing guidelines lack a shared definition of user trust in AI and its implementation requirements.

Abstract

Artificial intelligence (AI) is at the forefront of innovation in medicine. Researchers and AI developers have often claimed that "trust" is a critical determinant of the successful adoption of AI in medicine. Despite the pivotal role of trust and the emergence of an array of expert-informed guidelines on how to design and implement "trustworthy AI" in medicine, we found little common understanding across these guidelines on what constitutes user trust in AI and what the requirements are for its realization. In this article, we call for a conceptual framework of trust in health-related AI which is based not just on expert opinion, but first and foremost on sound empirical research and conceptual rigor. Only with a well-grounded and comprehensive understanding of the trust construct, we will be able to inform AI design and acceptance in medicine in a meaningful way.