Large Language Model Few-Shot Prompting with Dilemma Training Outperforms Human Surrogates in Predicting Patient Preferences
Authors: Natasha Ureyang, Sebastian Porsdam Mann, Yuxin Liu, Zuriel Hassirim, Melanie Almonte, Wenhao Chen, Joyce Ng, Thant Nay Lin, Aung Thiha, Gerald CH Koh, Brian David Earp, Pin Sym Foong
Organizations: Telehealth Core, National University of Singapore Singapore, Singapore · Centre for Advanced Studies in Bioscience Innovation Law, Faculty of Law University of Copenhagen Copenhagen, Denmark · Centre for Biomedical Ethics, National University of Singapore Singapore, Singapore · Imperial College London, Epidemiology and Biostatistics, School of Public Health London, United Kingdom
In serious illness, human surrogates often struggle to accurately predict patient preferences (68% accuracy), causing decision conflict. Personalized Patient Preference Predictor (P4) agents offer a potential solution, but prior prototypes treat values as static ratings, ignoring the contextual, situation-dependent nature of medical choices. Grounded in the 'logic of care', we present P4-DT (Dilemma Training), a P4 agent that constructs a patient decision policy by engaging users with varied medical dilemmas, eliciting individual preference reasoning through bi-directional training. In a study with 12 patient-surrogate dyads, P4-DT predicted patient treatment choices with 81.7% accuracy, significantly exceeding chance (OR = 5.61 [2.03, 15.51], p < .001) and outperforming both unassisted surrogates (55.0%; OR = 3.67 [1.59, 8.47], p = .002) and surrogates assisted by P4-DT (61.7%). Comparative prompt analyses showed that incorporating contextual scenario decisions and open-ended text improved accuracy by 15.0 percentage points over initial values ratings alone. We discuss implications for further testing and designing of context-aware AI agents that embody richer human experience to partner in complex decision-making.
Figures & tables
Figure 1. Accuracy of predicted patient preferences: P4-DT (AI agent), TO (human surrogate), and TO + P4-DT (human surrogate with P4-DT asisstance). On average, P4-DT predicted patient preferences with 81.7% accuracy compared to 55.0% for TO and 61.7% for P4-DT
Figure 2. Prompt variations Prompt variations
Appendix figures & tables3 assets
Supplementary material from the paper’s appendix.
Appendix
Figure 3Figure 4
Figure 3. Breakdown of accuracy for each participant dyad . A comparison between the prediction accuracy of P4-DT (AI agent), TO (human surrogate), and TO + P4-DT (human surrogate with P4-DT asisstance) across 12 participant dyads (SS1–SS12), ranked by descending P4-DT accuracy. On average, P4-DT predicted patient preferences with 81.7% accuracy compared to 55.0% for TO. P4-DT outperformed TO in 7 of 12 dyads, matched TO in 4 dyads, and was outperformed by TO only once.