AI Standardized Patient Improves Human Conversations in Advanced Cancer Care

TL;DR

This study presents SOPHIE, an AI-powered standardized patient platform that combines large language models, a lifelike avatar, and automated, 3E-based feedback to train serious illness communication in advanced cancer care. In a randomized trial with 51 health professionals and students, SOPHIE produced significantly larger improvements across Empower, Be Explicit, and Empathize than a reading-only control, with high inter-rater reliability and large effect sizes, demonstrating the potential of AI-driven asynchronous training to scale complex clinician communication skills. The work highlights the system’s scalable design, actionable feedback, and user-perceived impact, while noting areas for improvement in avatar realism and long-term skill retention. Together, these findings suggest AI-based simulation can meaningfully augment clinician education and potentially improve end-of-life conversations, with implications for broader domains requiring sensitive interpersonal communication.

Abstract

Serious illness communication (SIC) in end-of-life care faces challenges such as emotional stress, cultural barriers, and balancing hope with honesty. Despite its importance, one of the few available ways for clinicians to practice SIC is with standardized patients, which is expensive, time-consuming, and inflexible. In this paper, we present SOPHIE, an AI-powered standardized patient simulation and automated feedback system. SOPHIE combines large language models (LLMs), a lifelike virtual avatar, and automated, personalized feedback based on clinical literature to provide remote, on-demand SIC training. In a randomized control study with healthcare students and professionals, SOPHIE users demonstrated significant improvement across three critical SIC domains: Empathize, Be Explicit, and Empower. These results suggest that AI-driven tools can enhance complex interpersonal communication skills, offering scalable, accessible solutions to address a critical gap in clinician education.

Figures (16)

• Figure 1: Randomized controlled study design. Participants were randomly assigned (blue arrows) to either the Control group, which completed a reading module on 3E communication skills, or the SOPHIE group, which received interactive practice with SOPHIE and personalized feedback. All participants engaged in standardized patient (SP) conversations both before and after the intervention, with evaluations provided by the SP and four third-party (TP) raters.
• Figure 2: Histogram of Overall scores before (red) and after (blue) intervention for both control and SOPHIE groups (bin size $0.05$). The figure illustrates that the SOPHIE group showed a more pronounced rightward shift
• Figure 3: Comparison of communication skill improvement between the Control and SOPHIE groups. Bars represent the mean change ($\Delta$) in total score in 3E, with error bars indicating the 95% confidence interval. Scores are normalized by minimum and maximum score, i.e., a 0.10 improvement represents a 10% improvement in the maximum possible score. *, **, and *** indicates $p<0.05$, $p<0.01$, and $p<0.001$ respectively.
• Figure 4: SOPHIE user speech to avatar response generation flow. Whole pipeline passes in nearly real-time.
• Figure 7: A hybrid skill classifier that detects whether a statement from the user falls under the 3E skills .