
Living wills serve as an institutional link connecting medical self-determination, respect for life's dignity, family care ethics, and palliative care. China's existing law already provides a normative foundation through provisions on the right to life, informed consent, emergency rescue, voluntary guardianship, and medical quality management. These rules, however, remain fragmented and are insufficient to address the preservation of patients' authentic wishes, the boundaries of relatives' participation, the liability risks faced by medical institutions, and dispute resolution. A China-specific living will system should take the priority of the patient's own directive as its core rule and confine family involvement to a procedural mechanism for confirming, interpreting, and supporting the patient's wishes. Specifically, building on existing normative resources, the system should clarify the priority effect of living wills over medical decisions and limit their scope to major medical matters in specified serious life-threatening conditions; improve the rules on making, modifying, and revoking living wills; and safeguard implementation through physician review, ethics review, medical record documentation, and liability allocation. In this way, living wills can be embedded in China's existing health law system, yielding an institutional arrangement that is both locally adaptable and clinically operational.
When patients lack decision-making capacity, identifying, interpreting, and giving effect to their wishes remains a central challenge in surrogate decision-making. The personalized patient preference predictor (P4) seeks to support preference identification through large language models. However, patient preferences are closely tied to life experience, family relationships, and the care situation, and cannot be adequately understood from digital traces alone. In family-based surrogate decision-making in China, P4 may reproduce existing biases embedded in its training data; expanding the use of private data to improve accuracy may also exceed the scope of patient authorization and increase the risk of privacy violations. Its data use should therefore follow the principle of data minimization, its predictions should be interpreted in conjunction with clinical judgment and within family deliberation, and patients' clearly expressed prior wishes and best interests should serve as fundamental limits. Moreover, predictive accuracy does not by itself confer normative authority: over-reliance on algorithmic outputs may narrow the space for deliberation and obscure where responsibility lies. P4 should therefore be positioned as an accountable, contestable, and constrained input to family deliberation. Clarifying the responsibilities of the parties involved, establishing procedures for objection and review, and defining limits on use can help safeguard patient dignity and preserve the space for deliberation. Drawing on the norms of patient autonomy, substituted judgment, and the best-interests standard, this article analyzes the ethical dilemmas that arise when P4 is introduced into family-based surrogate decision-making in China, and examines its normative status and procedural safeguards.