AI in End-of-Life Decision-making: enhancing participatory agency through the Perfector Model

Artificial intelligence (AI) is increasingly influencing healthcare, including the ethically sensitive domain of end-of-life decision-making. A notable contribution in this area is Earp et al.’s Personalized Patient Preference Predictor (P4), an AI-based tool designed to infer the treatment preferences of incapacitated patients by analysing personal data. While intended to uphold patient autonomy, the model raises concerns regarding the role of algorithmic inference in capturing the complexity of individual values, and the risk of reducing deeply human-centred decisions to data-driven predictions. In response, we propose a conceptual alternative: the Personalized Patient Preference Perfector. This model shifts the emphasis from prediction to simulation and reflection. Rather than extracting or inferring preferences post hoc, the Perfector is designed to support patients with decisional capacity in articulating and exploring their values in the context of advance care planning. Drawing on the normative principles underpinning traditional advance directives, such as those defined in Italian Law 219/2017, it aims to enhance ethical dialogue between patients and healthcare professionals. Situating the Perfector within broader debates on autonomy, beneficence, and participatory care, we argue that such tools can be ethically valuable insofar as they foster morally meaningful engagement with end-of-life choices in an increasingly predictive and personalized medical landscape.