Beyond Prediction: The Ethical Limits of AI in Surrogate Decision-Making

Abstract: Artificial intelligence is transforming bioethics, yet its potential role in surrogate decision-making remains largely unexplored. If AI can create digital twins to model a patient’s physical health, could it also simulate their medical decision-making? This paper examines whether AI-driven models could serve as surrogates when patients lack advance directives or designated proxies.

The conventional substituted judgment standard prioritizes surrogates who best understand the patient’s preferences. If an AI model could more reliably reconstruct a patient’s decision-making than human surrogates, it may seem preferable. However, I argue that surrogacy is not just an epistemic task but also a normative one. Even if AI can accurately predict a patient’s choices, it fails to meet the moral criteria for surrogacy—namely, the capacity to bear and fulfill duties, particularly those related to human rights and social relationships.

This paper critiques the received view of surrogate decision-making and proposes a rights-based framework. I argue that surrogates permissibly act on behalf of incapacitated patients only if they serve as duty-bearers to the patient’s social rights. While AI may enhance epistemic reliability, it is unclear whether it can fulfill these moral obligations. If surrogacy requires more than knowledge—if it demands relational and ethical responsibilities—then AI, regardless of its predictive accuracy, cannot serve as a legitimate surrogate. This analysis challenges assumptions about AI’s role in medical ethics and underscores the enduring importance of human relationships in surrogate decision-making.

Keywords: AI, Surrogate decision-making, social rights