Ethical Algorithmic Advice: Some Reasons to Pause and Think Twice.

Abstract

Machine learning and other forms of artificial intelligence (AI) can improve parts of clinical decision making regarding the gathering and analysis of data, the detection of disease, and the provision of treatment recommendations. The target article “ Algorithms for Ethical Decision-Making in the Clinic: A Proof of Concept ” (Meier et al. 2022) explores the less-exam-ined possibility of using this technology to provide ethical advice. The article examines the feasibility of an algorithmic advisory system for clinical ethics called METHAD, which is designed to provide recommendations to clinicians facing difficult ethical ques-tions. METHAD utilizes a form of machine learning model called fuzzy cognitive maps and is based on Beauchamp and Childress ’ four principles of biomedical ethics, namely, beneficence, non-maleficence, autonomy, and justice, and is trained on data from clinical ethics committees. The article provides an illu-minating and highly interesting exploration of how ethical principles can be operationalized into an algorithmic model, which clinicians could use as an advisory tool and even defer to for moral judgments, similar to how they might defer to people concerning ethical issues. The authors also display a sensible degree of expert humility on behalf of METHAD and are explicit about the technical and ethical challenges regarding the reliability and acceptability of the recommendations that the algorithm provides. In this commentary, we wish to draw attention to some algorithmic pertaining ethical algorithmic design and the of