In narrow and discrete settings, AI therapy chatbots are proving effective means of delivering mental health interventions. 

A recent randomized-control trial study of Therabot by Dartmouth College researchers finds “significantly greater reductions of symptoms” for major depressive disorder (MDD), generalized anxiety disorder (GAD) and those at high risk for eating disorders. 

“Fine-tuned Gen-AI chatbots offer a feasible approach to delivering personalized mental health interventions at scale, although further research with larger clinical samples is needed to confirm their effectiveness and generalizability,” the study states.

A press release from Dartmouth says that trial participants felt they could trust the therapy chatbot to a degree that was comparable to working with a real therapist. 

The intervention group totaled 106 people with the above-mentioned clinical presentations. They used Therabot via smartphone app for an average of more than 6 hours, roughly equivalent to 8 45-minute therapy sessions, over eight weeks.

By the end of the trial period, the Therabot group saw symptom reductions greater than the control group. The GAD and eating disorder groups saw about 65% better symptom scores, with the MDD seeing 47% better scores

About three-fourths of the Therabot group were not under pharmaceutical or other therapeutic treatment at the time, according to the release.

“Our results are comparable to what we would see for people with access to gold-standard cognitive therapy with outpatient providers,” Nicholas Jacobson, the study’s senior author and an associate professor of biomedical data science and psychiatry at the Geisel School of Medicine, said in the release. “We’re talking about potentially giving people the equivalent of the best treatment you can get in the care system over shorter periods of time.”

Additionally, therapeutic alliance, one of the few universally accepted indicators of care quality and likelihood of success, was in line with working with a real therapist, according to the release.

Jacobson also said in the release that there is “replacement for in-person care,” despite the study demonstrating equivalence between the two interventions, and also said in the release that “there are nowhere near enough providers to go around.”

“I see the potential for person-to-person and software-based therapy to work together,” Jacobson said.

Michael Heinz, the study’s first author and an assistant professor of psychiatry at the Dartmouth College Center for Technology and Behavioral Health and the Geisel School of Medicine, said in the release that “no generative AI agent is ready to operate fully autonomously in mental health.”

He highlighted the range of risk scenarios prevalent in outpatient mental health, the lack of understanding around the risk of generative AI models.

Therabot’s development began with the Dartmouth AIM HIGH Lab, led by Jacobson, in 2019.

“There are a lot of folks rushing into this space since the release of ChatGPT, and it’s easy to put out a proof of concept that looks great at first glance, but the safety and efficacy [are] not well established,” Jacobson said in the release. “This is one of those cases where diligent oversight is needed, and providing that really sets us apart in this space.”

For the most part, AI tools that are in the market and being used by providers are largely focused on administrative matters such as note-taking and data gathering. Therapy chatbots have been seen skeptically by most providers.