News Release

Artificial Intelligence for Mental Health?


A recent STAT newsletter reports that Woebot Health, a mental health company that uses artificial intelligence, is testing a new functionality using large language models (LLM) that interpret and generate text. The company has launched a two-week randomized trial of 150 participants to “gauge user satisfaction.” Some psychotherapists question both Woebot’s trial and STAT’s coverage of it.

    Michaels is a psychologist in private practice in Chicago and a co-founder of the Psychotherapy Action Network. 

Michaels told the Institute for Public Accuracy: “User satisfaction is a fairly meaningless metric to use if you want to measure treatment effectiveness. People can be ‘satisfied’ with what their therapist said, or even like their therapist, but those are very different things than what goes into making therapy effective. On the other hand, if you want to build a product and sell it and create customers and make money, then you might want to measure user satisfaction.

“A two-week time frame also might mean something in the world of for-profit app development, but it’s meaningless in terms of therapy, therapeutic value, and how people grow and change. 

“It sounds like their goal is clear: a shorter path to commercial viability. This is, once again, a fundamental mismatch between the goals of a for-profit technology business and the aims of a therapeutic endeavor. The therapeutic endeavor relies to a large extent on the relationship between two humans. Any technology tools trying to enter this therapy space need to deeply understand how therapy works, how people change, and then from there, develop a plan for what their tool might do, and accept and appreciate what it cannot do.

“Instead of fiddling around the edges and trying to leverage LLMs to make its responses sound better, I’d like to see Woebot start from a deep understanding of the potential benefits and limits of their tool—and build something that is useful to helping people, or supporting people’s efforts to grow and learn and change.” 

Michaels is part of a group of academics from the U.S. and the U.K. researching digital therapy tools, called the Digital Therapy Project. The group’s first survey recently launched; it focuses on documenting the landscape of digital therapeutics. The group, Michaels said, “aims to help identify which tools and technology might be helpful in improving [patients’] coping mechanisms, mental health, self awareness, and self understanding.”