Problems with a one-size-fits-all approach: a systematic literature review on solutions to AI bias in healthcare and implications for engineering biology

Much of the current literature on artificial intelligence bias in healthcare presents a one-size-fits-all approach to bias mitigation. Such approaches, however, are unlikely to offer actionable guidance to those developers (and their institutions) who are looking for strategies to minimise potential bias. This paper presents findings from a systematic literature review exploring how AI developers can mitigate bias in the field of engineering biology. A systematic search was conducted on the Scopus and Web of Science databases for relevant articles published between 2015 and 2024. The findings from 51 reviewed articles show that recommendations for bias mitigation within healthcare tend to be grouped around seven key themes, namely diversity in teams, training and education, awareness and responsibility, diversity of data, collaboration with end users and beneficiaries, monitoring and evaluation, and transparency. While these recommendations provide useful suggestions for reducing bias, they generally fail to provide actionable guidance or empirical evidence about how these strategies can be operationalised in a real-world setting. More research is needed to test the effectiveness and practicality of these recommendations across different scientific and clinical contexts as well as among different types of development teams.