Abstract

Medical decision-making on behalf of individuals who have lost the capacity to make their own treatment choices poses significant challenges. The “substituted judgment” standard, which prioritizes individual autonomy, requires surrogate decision-makers to choose the course of action that the patient would have chosen or endorsed if they were able to do so. However, research suggests that surrogates often face difficulties in accurately predicting patient preferences, even when making a good-faith effort to adhere to the substituted judgment standard. To address these challenges, the Patient Preference Predictor (P3), a computer-based algorithm that correlates demographic data with expressed medical preferences, has been proposed to supplement or replace aspects of the surrogate decision-making process. However, critics argue that the P3 fails to respect patient autonomy due to relying on population-level data rather than accounting for individuals’ unique values, beliefs, and preferences. In response, we have proposed with our colleagues the Personalized Patient Preference Predictor (P4), an artificially intelligent (AI) system that would use machine learning to create a “digital psychological twin” or AI simulation of a patient, inclusive of their values, beliefs, and preferences, from individual-level data. While the P4 aims to address concerns raised by the P3, it has been argued that it may introduce new problems, particularly regarding its potential negative effects on the doctor–patient relationship and family relationships. In this chapter, we describe the P4, summarize objections related to its impact on human relationships, respond to these objections, and conclude with some thoughts on future directions.