In June 2017, the journal JMIR Mental Health published “Delivering Cognitive Behavior Therapy to Young Adults With Symptoms of Depression and Anxiety Using a Fully Automated Conversational Agent (Woebot): A Randomized Controlled Trial” by Kathleen Kara Fitzpatrick, Alison Darcy, and Molly Vierhile. It was an early attempt to subject a companion-style mental-health chatbot to the standard of evidence used for medical interventions.
The study randomized 70 college-aged participants (ages 18 to 28). One group spent two weeks talking with Woebot, a fully automated chatbot that delivered cognitive behavioral therapy (CBT) content in a conversational format; the control group was directed to a National Institute of Mental Health e-book on depression. After two weeks, the Woebot group showed a statistically significant reduction in depression symptoms on the PHQ-9 scale relative to the information-only control (F=6.47, P=.01), a moderate between-groups effect. Participants engaged frequently, and qualitative feedback attributed much of the benefit to the bot’s perceived empathy and personality.
The trial mattered less for its size - it was small and short - than for its framing. It treated a chatbot not as a novelty but as a clinical tool that should be measured, and it foregrounded the same finding ELIZA’s users had stumbled into in 1966: people will open up to a program that listens, and that opening can be put to therapeutic use. Woebot, founded by Darcy, became one of the most studied digital-health chatbots of the following decade.
Why business readers should care: the line between a wellness chatbot and a medical device is exactly the willingness to run trials like this one - claims about mental-health benefit invite the regulatory and liability scrutiny that come with treating, not just entertaining, a user.