Artificial intelligence and medical futility

BACKGROUND: Determining when medical treatment is futile is conceptually, empirically, and ethically disputed. This creates challenges for patients, families, doctors, healthcare, and legal systems. Amid such disputes, there is still a practical need to resolve futility questions in individual cases. Can artificial intelligence (AI) be ethically and effectively used to help with such decisions? METHODS: We adopted a critical narrative review method. We first surveyed existing scholarship on the ethics of medical futility to identify the main positions in ongoing futility debates. We then surveyed existing or potential medical AI devices to identify whether such tools might be used to help address these debates or their practical upshots. Finally, we philosophically analysed whether such tools ought to be used, identifying and weighing reasons using analytic argument, applying relevant ethical concepts, and philosophical principles, such as respect for autonomy, beneficence, non-maleficence, and justice. RESULTS: We found that there are some key challenges or risks in using AI for such purposes, for example, using AI may exacerbate the problem of self-fulfilling prophecies in futility determination by absorbing predicted outcomes as data. Nevertheless, under certain conditions, medical AI could ethically be used to prevent or help resolve futility disputes in healthcare decision-making. CONCLUSIONS: AI could ethically contribute to preventing or resolving futility disputes depending on how AI is integrated and regulated in end-of-life care.