top of page

Rise of the robot therapists

Meet your robot therapist. He's not as revolutionary as you might think. Ever had online customer service with a chatbot? Maybe you even thought it was a real person. And, "teletherapy," using online videochat with a real therapist, is increasing radically in popularity. Combine the two, and you have Woebot, who dispenses cheap and apparently effective Cognitive Behavioral Therapy (CBT). This post is a reaction to a study of Woebot with depressed and anxious college students.

Some disclaimers, before I get into analysis:

1) Woebot does not screen for trauma/abuse, which I find is behind over half of the anxiety and depression cases I see.

2) Woebot does not screen for trauma/abuse, which I find is behind over half of the anxiety and depression cases I see. Cognitive therapists in such cases are treating symptoms, and most often, are fighting a losing battle. Trauma needs trauma therapy.

3) I do not like teletherapy. It was not given the green light by my professional organization until a year ago due to ethical concerns I still hold, such as the fact that nothing on the internet is private--nothing. More importantly, if a face on a screen were as good as a real person for someone in trouble, why would soldiers care about being away from their families if they can Skype? If the best you can say about a therapy method is that "it's better than nothing," that's pretty pathetic, and a good indicator that the real issue is getting more people into real therapy.

4) I am not endorsing the use of Woebot, and I think he needs much more research to be sure he's safe.

This said, Woebot got some pretty good initial results. It retained "clients" better than most therapists (90%), which at least shows that people found it helpful, and that it may have prevented depression from worsening. More importantly, 100% of "clients" reported discovering something new, and talking to Woebot about problems every couple days over 2-3 weeks reduced depression on the PHQ-9 survey of depressive symptoms (about 2.5 points on a 27-point scale). This result was statistically significant, meaning that the regular ups and downs people with depression experience were accounted for. Interestingly, most of the benefit was in physical symptoms of depression, like having increased energy and regaining appetite. I use the PHQ-9 on a regular basis, so this finding hit home with me. The graph below shows overall reduction in depression by participants using Woebot, compared to the control group (who were given a self-help book).

What I would like to see is a 3-month comparison of Woebot to human Cognitive Behavioral (CBT) therapists. CBT is a therapy that can be done by-the-book, for the most part, and largely relies on relationship, discussion, and homework. So it's easier to teach to a robot. I'll bet the robot is more effective than the therapists, and I'll tell you why. One of the big problems with depression is that people get stuck at home and get into bad habits of not taking care of themselves. Woebot comes into the home with no appointment needed, and will buzz your phone just to see if you want to talk for 5 minutes. I think this is why it was the most effective with physical symptoms of depression, and that to me is huge. So often, depressed people know what they should do to fix the situation without a therapist arguing them into it. They just have an overwhelming tiredness and inertia that undercuts any insight gained in session. They have an amazing session, but don't do the homework, and that's the definition of a therapy epic fail. But Woebot, as I mentioned, undercuts the physical inertia that doesn't respond to logic, and that foothold I think would make him climb higher in the long term than a human, in bringing insights to fruition. Someone motivated to be active (for instance by Woebot, apparently) can do things that reinforce a desire to live, such as get friends and get a job, frankly. But developing a desire to live, as awesome as that is, has a tougher time motivating people out of bad living habits, which leads to that epic fail I mentioned, and a return to depression.

Am I for or against Woebot? I think there may be a place for him to keep therapists honest. There is so much bad therapy out there, that this study merits a long-term comparison to human therapists. My main issue, again, is that it doesn't screen for trauma. I would feel more comfortable if it began with that, and if it were more of a stepping-stone into therapy. Ideally, Woebot could help depressed people define what they want out of life, and motivate them physically to get out and chase that--and then work on difficult emotions and life experiences with a real therapist, without so much risk of dropping out of therapy for lack of progress due to motivation issues.


*UPDATE 12/23/17*

A client of mine reported having used Woebot for one week, and found him "woe-fully" lacking in the area of artificial intelligence. The client found that Woebot could dispense a lot of advice, but his reactions to simple back-and-forth chat were awkward and repetitive.


The full article is interesting as it goes into more detail about the effects Woebot's quirky, attractive personality had on participants. [Click here for full article]

If you are really interested and can skim the more boring parts discussing statistics you may not be interested in (although stats are the real way to cheat in scientific research), you can read the study here: [Click here for full study]

bottom of page