Advancing empathy in AI: Vanderbilt tackles chatbot safety for mental health

Artificial intelligence has become a valuable part of many areas of health care. A team at Vanderbilt University Medical Center is setting out on a two-year project that aims to make AI chatbots more effective — and more empathetic — when part of behavioral health care treatment.

The project will focus on major depression and generalized anxiety disorder. “From a safety and reliability standpoint, these chatbots pose a number of lingering challenges and unresolved questions,” said Susannah Rose, associate professor of biomedical informatics and the principal investigator for the project. By combining human expertise with computational techniques, the team hopes to create a chatbot that can detect hallucinations, omissions and misaligned values. The end goal is for the chatbot to be able to provide AI-generated responses that adapt to diverse populations, thereby becoming more empathetic to the patients it will serve.

“By incorporating community members, patients and clinicians throughout the process, we aim to create a system that not only improves technical accuracy but also aligns with users’ diverse values and expectations,” Rose said.