A Scoping Review of Algorithmic and Data-Driven Technology in Online Mental Healthcare: What is Underway and What Place for Ethics and Law? (Preprint)
BACKGROUND Uncertainty surrounds the ethical and legal implications of algorithmic and data-driven technologies in the mental health context, including technologies variously characterised as artificial intelligence, machine learning and deep learning. OBJECTIVE We aimed to survey scholarly literature on algorithmic and data-driven technologies used in online mental health interventions with a view to identify the legal and ethical issues raised. METHODS We searched for peer-reviewed literature about algorithmic decision systems in mental healthcare used in online platforms. Scopus, Embase and ACM were searched. 1078 relevant peer-reviewed research studies were identified, which were narrowed to 132 empirical research papers for review based on selection criteria. We thematically analysed the papers to address our aims. RESULTS We grouped the findings into five categories of technology: social media (n=53), smartphones (n=37), sensing technology (n=20), chatbots (n=5), and other/miscellaneous (n=17). Most initiatives were directed toward “detection and diagnosis”. Most papers discussed privacy, principally in terms of respecting research participants” privacy, with relatively little discussion of privacy in context. A small number of studies discussed ethics as an explicit category of concern (n=19). Legal issues were not substantively discussed in any studies, though seven studies noted some legal issues in passing, such as the rights of user-subjects and compliance with relevant privacy and data protection law. CONCLUSIONS Ethics tend not to be explicitly addressed in the broad scholarship on algorithmic and data-driven technologies in online mental health initiatives—even less so legal issues. Scholars may have considered ethical or legal matters at the ethics committee/institutional review board stage of their empirical research but this consideration seldom appears in published material in any detail. We identify several concerns, including the near complete lack of involvement of service users, the scant consideration of ‘algorithmic accountability’, and the potential for over-medicalisation and techno-solutionism. Most papers were published in the computer science field at a pilot or exploratory stage. Thus, these technologies could be appropriated into practice in rarely acknowledged ways, with serious legal and ethical implications.