Suppression of algorithm aversion through modifying AI’s decision in an interpretation of radiogram task

Abstract

In today's society, the use of Artificial Intelligence (AI) has become a part of our daily lives, and collaboration with AI is increasingly important. However, there's a phenomenon where people tend to prioritize human-generated results over those generated by AI, despite AI's superior capabilities. This phenomenon is called "algorithm aversion". Individuals with high levels of expertise, such as radiologists who require specialized skills, often exhibit a greater tendency towards algorithm aversion, namely they may not rely on results generated by algorithms. Previous research has suggested that even slight adjustments to AI outputs could potentially reduce algorithm aversion. Against this background, the primary goal of this study is to evaluate and mitigate algorithm aversion in the context of X-ray image interpretation by modifying AI’s decision. Specifically, we aim to investigate whether individuals are more likely to rely on and choose AI diagnoses when they have the option to modify AI’s decision. It is anticipated that this intervention will serve to mitigate algorithm aversion exhibited by radiologists and consequently foster enhanced collaboration with AI.