Psychiatric Advance Directives and Artificial Intelligence: A Conceptual Framework for Theoretical and Ethical Principles

Abstract

The patient's decision-making abilities are often altered in psychiatric disorders. The legal framework of psychiatric advance directives (PADs) has been made to provide care to patients in these situations while respecting their free and informed consent. The implementation of artificial intelligence (AI) within Clinical Decision Support Systems (CDSS) may result in improvements for complex decisions that are often made in situations covered by PADs. Still, it raises theoretical and ethical issues this paper aims to address. First, it goes through every level of possible intervention of AI in the PAD drafting process, beginning with what data sources it could access and if its data processing competencies should be limited, then treating of the opportune moments it should be used and its place in the contractual relationship between each party (patient, caregivers, and trusted person). Second, it focuses on ethical principles and how these principles, whether they are medical principles (autonomy, beneficence, non-maleficence, justice) applied to AI or AI principles (loyalty and vigilance) applied to medicine, should be taken into account in the future of the PAD drafting process. Some general guidelines are proposed in conclusion: AI must remain a decision support system as a partner of each party of the PAD contract; patients should be able to choose a personalized type of AI intervention or no AI intervention at all; they should stay informed, i.e., understand the functioning and relevance of AI thanks to educational programs; finally, a committee should be created for ensuring the principle of vigilance by auditing these new tools in terms of successes, failures, security, and relevance.

Keywords: artificial intelligence; clinical decision support system; joint crisis plan; medical ethics; predictive medicine; psychiatric advance directives.

Figure 1

Patient's preference seen as a multivariate function, here represented as a function of time. Real conscient/public (blue line) and subconscient/private (brown line) preferences variate with time. Psychiatric advance directives (PADs) only gives the conscient/public preference but shifted in time (blue dotted line) since there is a time dt between the PAD drafting and the moment it applies. Artificial intelligence (AI) implemented to PADs infers conscient/public preference shifted in time but also shifted in nature (dashed blue line) by definition of inference: it do not gives the real preference anymore. AI could also infer subconscient/private preference (dashed brown line).

Figure 2

Potential places of artificial intelligence (AI) in the contractual relationship during the psychiatric advance directive (PAD) drafting process. PADs can be considered as a third-party contract in a tripartite relationship between the patient, the care team, and a third-party such as the family or the person of trust. AI could act as a substitute of a party (A), as a fourth-party (B), or as a partner to each party (C).