Explainable AI for Breast Cancer Diagnosis: Application and User’s Understandability Perception

Abstract

With the current progress of how Artificial Intelligence (AI) is implemented in different fields, the concerns on AI technologies’ accountability, transparency, trust, and social acceptability are also raised. These concerns become even bigger when people’s well being is at stake, as in the case of AI applications implemented in healthcare. Explainable AI (XAI) or AI explanation algorithms are proposed to solve the accountability and transparency problems. However, how users perceived algorithm explanation have not yet been explored extensively. In this paper, Explainable AI approaches were implemented to the specific case in healthcare: Breast Cancer. An online survey was conducted to investigate users’ perception and understanding of the AI explanation algorithms: LIME and Anchors. We were looking at the users’ perception of Explainable AI, specifically non-expert users, and found that users’ perceived understanding was high even though the majority did not understand the explanation.