I often get asked for suggestions as to which mental health apps work best. These days, I also get the occasional request for AI-powered therapy. But what do we know so far about the effectiveness of AI therapy apps?
While peer reviewed studies are essential, users’ qualitative experience are also crucial in evaluating the effectiveness of AI therapy apps. One app that uses AI-powered conversation is Woebot. Writing on Very Well Mind, Sarah Bradley gives a very helpful overview of her experience with the app.
One user’s experience of Woebot
The app combines psychology research with AI technology, Sarah Bradley explains. And offers cognitive behavioural therapy (CBT) techniques through conversations and educational modules. She tested the app for 30 days alongside her regular therapy sessions. And her write-up explores the effectiveness of this AI therapy app in helping her manage her anxiety and OCD symptoms.
The app was free, easily accessible, and provided some helpful insights. But she felt restricted by the app’s scripted responses. And found it unable to effectively address complex emotions or provide personalised support. Ultimately she found it lacking compared to traditional therapy.
She suggests that Woebot may help individuals with limited mental health issues or no prior therapy experience. But also that it may not be suitable for those seeking deeper or more tailored support. In particular, it cannot replace human therapists, especially for clients with more significant mental health concerns or prior therapy experience.
The relevance of the computer-human bond
Decades of research shows that the therapist-client relationship is the single most impactful factor in determining the effectiveness of therapy. On this point, a formal (2021) study of Woebot’s use of “digital conversational agents” to bond with humans is interesting. The study’s first-named author is Alison Darcy, the company’s founder and president. Of course she has a stake in Woebot’s success, but her bio is clearly not that of a digital wannabe looking for a fast buck. So we can respect the fact that the study points only vaguely to future potential. It would be interesting to see a more recent follow-up.
The future effectiveness of AI therapy apps
It is reasonable to expect the conversational abilities of therapy bots to improve greatly as the computing power behind large language models increases rapidly. And it remains to be seen whether this will translate into a computer-human bond good enough to underpin effective psychotherapy. For now, I am happy to agree with Sarah Bradley: my job is not under imminent threat from such apps!
***
Header image by Delta Works.