A Scoping Review of Health AI Controversies in the Grey Literature from 2013-2022

Objective This study aimed to 1) quantify the number of controversies related to health artificial intelligence (AI) development and deployment over the past 10 years, and 2) categorize health AI controversies by theme. Materials and Methods This study is a scoping review. We queried GoogleNews for articles meeting pre-specified inclusion and exclusion criteria related to health AI controversies. A total of 7508 articles were queried, with 76 articles ultimately meeting criteria. Articles were quantitatively evaluated for timing and coded according to both the organization and the themes of the controversies described in the article. Results Of the 76 articles, 55% were published in 2019 and 2022. No articles were published before 2016. Privacy, AI accuracy, and AI bias were the most common themes. Oversight and conflict of interest were the least common themes. Google, Babylon, and Optum were the most discussed companies. Conclusions The results demonstrate concrete evidence for many of the theoretical concerns highlighted by patient and physician surveys as well as scholarly ethics research. However, the volume of articles in the 10-year study period was small. It is possible that this study was too narrow in scope, that health AI has actually not yielded a large volume of controversial events, or that news organizations have insufficiently investigated potential health AI controversies. Further research is required to develop a more complete understanding of controversial activities in health AI.