Stanford Medicine is piloting an artificial intelligence platform that lets physicians rehearse challenging conversations with virtual patients, in an effort to strengthen communication skills without adding to already packed schedules.
The tool, called the Care Collaborative Professional Empathy Training Platform, or Care Collab, was developed by Flora Ma, PhD, of Stanford Health Care’s Department of Psychiatry and Behavioral Sciences.
The goal is straightforward: give physicians a low-stakes way to practice real-time conversations — including those that may involve strong emotions, resistance to treatment, or complex decisions — using AI-driven patient simulations built on ACES’ evidence-based communication framework.
“Doctors have been struggling with having difficult conversations with diverse patient populations,” Ma said. “They don’t always get real-time training and feedback on interactions. The technology is supposed to simulate scenarios that allow them to practice communication skills and also serve as a reflection point for them.”
The tool will be integrated into Stanford’s existing Docebo learning management system as part of the Advancing Communication Excellence at Stanford (ACES) program.
Mike McAuliffe, a program manager with the TDS Program Management Office who supports the Stanford EdTech team, said he became involved when the Care Collab team sought to host the platform on Docebo, one of the learning management systems EdTech supports.
“One of the primary use case scenarios for AI that the EdTech team's been receiving consult requests on over the past year has been finding ways to provide more of this asynchronous, very low-stakes communication practice opportunity,” said McAuliffe.
“Trying to practice these conversations person to person requires a level of vulnerability — that oftentimes people who are experts in their field have a little bit of trouble with. And so being able to give people these low-stakes opportunities to practice via AI whenever they want, however many times as they want, just to stretch those muscles, has been one of the key areas EdTech’s been consulting in.
“While this isn’t an EdTech project, we're very excited to see this go live and to have more of these approaches out in our community so that we can all learn and share together.”
Unlike traditional online modules that rely on multiple-choice questions, Care Collab presents a video-based, conversational interface. The physician speaks with an AI “patient” while the system records both verbal and nonverbal behavior, then scores the interaction against specific competencies, such as asking open-ended questions or responding to emotion.
If a physician asks, “What are your thoughts about starting this medication?” the platform can flag that moment in the video and transcript as evidence of an open-ended question. Those data points are mapped to the ACES framework, giving users concrete examples of what they did well and where they might adjust their approach.
Stephanie Chen, Manager for Patient Experience Programs, noted, “A lot of our physicians want one-on-one coaching, and that requires a lot of time. So having the ability to integrate Care Collab into our inpatient care course is going to be amazing, because they'll be able to practice in real time over and over again for as many times as they want. And it's going to be customized to their specialty.”
The first planned use of Care Collab will be in an ACES course for inpatient care, aimed at about 100 hospital medicine physicians in its initial year, with broader rollout anticipated later. The course is designed to take about an hour and can be completed asynchronously on a physician’s own device. Participants can return to the platform to try additional scenarios or revisit skills over time.
“We heard from our physicians that the most feasible version of a training would be one hour, asynchronous, something they could access from their own device,” said Rachyl Pines, PhD, a communication researcher who leads the ACES team. “We also know that role play and facilitated discussion are really useful. This tool is a place where physicians can enact the new skills they’re learning without having to schedule a time to be live with a trainer.”
The platform is configurable: departments can tailor patient personas, scenarios and grading criteria to their own needs. Access to recordings and transcripts is controlled through role-based permissions, allowing programs to decide how much faculty or supervisors can see.
If the pilot launches as planned in the new fiscal year, Care Collab could eventually extend beyond hospital physicians to residents, nurses and other clinicians — a broader audience that, Ma notes, often spends even more time at the bedside.
Beyond training, Pines sees research potential. Because the system captures detailed behavioral data, it may offer a more precise view of how communication skills develop over time than post-visit surveys or interviews typically allow. “I think it's so cool that we’re thinking about ways to include emerging technologies to continue to enhance communication skills,” she said.
