Johns Hopkins psychiatrist Jacob Taylor wants to use artificial intelligence to give more patients access to psychiatric care.

Taylor, with assistant professor Ananya Joshi and postdoctoral researcher Synthia Wang, is working on a tool called PhippsBot that will gather information from patients through AI-powered text chats in order to make intake interviews more efficient.

“Typically, those interviews can take several hours, and they require a sensitivity to the nuances of the information being conveyed,” Taylor says. “You might ask a person a question and maybe they misunderstood the intent. Or you notice that something they said just now seems inconsistent with something they said 15 minutes ago. You have to probe and tease out that inconsistency.”

PhippsBot would take over some of that probing and generate a report for the psychiatrist to read before the interview, saving time and reducing the bottleneck that keeps new patients from getting those initial appointments. The project is one of several in the Department of Psychiatry and Behavioral Sciences related to developing AI tools that could become models for the profession.

Joshi, who holds appointments in both psychiatry and computer science, leads the PALS (Psychiatry-Assistive Learning Systems) Lab, which develops AI systems for psychiatric care with a particular focus on rigorous evaluation, patient safety and clinical translation.

For PhippsBot, her lab leads the technical development and evaluation, working with Taylor and other clinicians so that it reflects the complexities of psychiatric intake. She notes that projects like PhippsBot must incorporate the high standards and expertise of Johns Hopkins Medicine. “What makes our psychiatrists special is that they notice the small things and know when to poke on them and when to pull back,” she says.

“As a field, we’re trying to figure out what aspects of a psychiatrist’s job AI can do and what aspects humans will always be necessary for,” says Taylor, who, as director of the residency program, helps train psychiatrists.

PhippsBot, he says, “is designed to assist your doctor in gathering information. It’s not giving advice. It’s not making decisions. It’s not prescribing medication. It’s not functioning as a therapist.”

Of course, AI is already being used in psychiatry; for more than a year, clinicians across Johns Hopkins Medicine have been using AI-powered transcription tools. They say it saves time and lets them focus more on patients, but assistant professor Roy Adams, also appointed in psychiatry and computer science, may be the first to ask what the technology means for patients.

Adams and co-investigators Joshi, Kevin Li and Peter Zandi in psychiatry, along with computer scientists Anjalie Field and Kristina Gligorić, recently won a $150,000 Johns Hopkins Discovery Award to analyze how these transcription tools may influence documentation of mental health symptoms and social determinants of health in emergency rooms and in primary care, psychiatric and Gyn/Ob clinics.

“A lot of the deployment of these tools into care is premised on the idea that they will either not impact care in a substantive way or they’ll improve care, but we just don’t have an evidence base for that,” says Adams.