Since 2013, Ieso has focused on depression and generalized anxiety disorders and has used data-driven techniques – NLP is a central part of which – to dramatically increase recovery rates for these disorders. According to Ieso, his 2021 recovery rate for depression is 62% – compared to a national average of 50% – and 73% for generalized anxiety disorder – compared to a national average of 58%.
Ieso says it focused on anxiety and depression, in part because they’re two of the most common conditions. But they also respond better to CBT than others, such as B. Obsessive-compulsive disorder. It is not yet clear how far the clinic can take its success, but it plans to focus on more conditions.
In theory, using AI to monitor quality means clinicians can see more clients, as better therapy means fewer unproductive sessions, although Ieso has not yet explored the direct impact of NLP on care efficiency.
“At the moment we can treat between 80 and 90 clients with 1,000 hours of therapy,” says Freer. “We try to move the needle and ask: Can you treat 200, 300, even 400 clients with it? Therapy hours? “
In contrast to Ieso, Lyssn itself does not offer any therapy. Instead, it makes its software available to other clinics and universities in the UK and US for quality control and training.
Lyssn’s US customers include a telemedicine opioid treatment program in California that seeks to monitor the quality of care provided by its providers. The company is also working with the University of Pennsylvania to use its technology to set up CBT therapists across Philadelphia.
In the UK, Lyssn works with three organizations, including the Trent Psychological Therapies Service, an independent clinic that – like Ieso – is contracted by the NHS to provide psychological care. Trent PTS is still testing the software. Since the NLP model was built in the US, the clinic had to work with Lyssn to identify British regional accents.
Dean Repper, Trent PTS Clinical Services Director, believes the software could help therapists standardize best practices. “You’d think that therapists who had been doing this for years would get the best results,” he says. “But they don’t necessarily do it.” Repper compares it to driving a car: “When you learn to drive, you learn some safe things,” he says. “But after a while you stop doing some of those safe things and you might get fined for speeding.”
Improve, not replace
The point of AI is to improve, not replace, human care. The lack of high quality mental health care will not be addressed by short term quick fixes. Addressing this issue also requires reducing stigmatization, increasing funding and improving education. Blackwell, in particular, rejects many of the claims made about AI. “The hype is dangerous,” he says.
For example, there has been a lot of buzz over things like chatbot therapists and around-the-clock monitoring through apps – often billed as Fitbits for the mind. But most of these technologies fall somewhere between “years away” and “never will happen”.
“It’s not about wellness apps and the like,” says Blackwell. “Giving someone an app that says they’ll treat their depression is probably just to vaccinate them from seeking help.”
One problem with evidence-based psychotherapy, however, is that therapists and clients are asked to open their private conversations. Will therapists object to such monitoring of their professional performance?
Repper expects a certain reluctance. “This technology presents a challenge for therapists,” he says. “It’s like having someone else in the room transcribing everything they say for the first time.” Initially, Trent PTS only uses Lyssn’s software on trainees who expect supervision. When these therapists qualify, Repper thinks, they can accept surveillance because they are used to it. More experienced therapists may need to be convinced of its benefits.
It’s not about using the technology as a stick, but as a support, says Imel, who was a therapist himself. He believes many will appreciate the additional information. “It’s hard to be alone with your customers,” he says. “If you’re only sitting in a private room with someone else for 20 or 30 hours a week without getting feedback from colleagues, it can be really difficult to improve.”
Freer agrees. At Ieso, therapists discuss the AI-generated feedback with their superiors. The idea is to let therapists take control of their professional development and show them what they’re good at – things that other therapists can learn from – that they may not want to work as well at.
Ieso and Lyssn are just beginning this journey, but there is clear potential to learn things about therapy that can only be revealed by obtaining sufficiently large data sets. Atkins mentions a meta-analysis published in 2018 that summarized around 1,000 hours of therapy without the aid of AI. “Lyssn processes that in a day,” he says. New studies published by both Ieso and Lyssn analyze tens of thousands of sessions.
For example, in a paper published in JAMA Psychiatry in 2019, Ieso researchers described a deep learning NLP model that was trained to categorize utterances by therapists in more than 90,000 hours of CBT sessions with around 14,000 clients. The algorithm learned to recognize whether different sentences and short segments of the conversation were instances of certain types of CBT-based conversation – such as checking the client’s mood, doing homework and checking (where clients practice the skills learned in a session), methods of change discuss, future planning, etc. – or conversations that have nothing to do with CBT, such as general chat.