AI preference prediction beyond substituted judgement: enhancing best interest decision-making

Area of Science:

• Medical Ethics
• Artificial Intelligence in Healthcare
• Clinical Decision-Making

Background:

• Tracking patient preferences is crucial for medical decisions, but current methods for incapacitated patients are often inaccurate.
• Surrogate decision-making for incapacitated patients faces challenges with accuracy and consistency.

Purpose of the Study:

• To evaluate the potential value of artificial intelligence preference predictors (AIPPs) in improving patient preference tracking for incapacitated individuals.
• To address the objection that AIPPs rely on impersonal information and demonstrate their utility in best interest decision-making (BIDM).

Main Methods:

• The study theoretically explores how AIPPs can support BIDM, focusing on improvements in accuracy, consistency, and speed.
• It analyzes the implications of AIPPs within the context of intensive care units (ICUs), where BIDM is prevalent.

Main Results:

• Even under strong objections regarding impersonal data, AIPPs can significantly enhance BIDM.
• The application of AIPPs in ICUs has substantial moral and practical consequences, offering a 'safe harbor' for further development.

Conclusions:

• AIPPs present a promising avenue for improving BIDM, enhancing accuracy, consistency, and speed.
• Further research is needed to address empirical and technical questions, establishing a clear agenda for AIPPs in determining best interests.