I am just about to read a news article (New York Times, June 1) about a “therapist” app named “Woebot.” I have read enough in the area that I think I have a sense of what has to be said. I don’t know the big names in the field, with the exception of Sherry Turkle, author of Alone Together, and I don’t know any studies at all. Still, it seems to me that whoever is talking, there are some things that have to be said.
This is what I think they are.
On the question of whether this is really “therapy,” the therapists  will be facing off against the technologists.  The therapists will ask if this is really therapy. Therapy is defined as a set of interactions that require human beings. Therapy requires honesty and insight; it requires empathy and understanding. Having declared that, they will turn to the technologists and dare them to show that Woebot has those things.
The technologists won’t do that, of course. They will change the presuppositions. “What is therapy, really?” they will ask. You guys are talking about history. This is the way it has always been done. Let’s talk, instead, about “it.” What is it, really, that needs to get done? People who are confused need to make a choice; people who need to change their behavior need a reason to do so and encouragement to continue, Woebot does that. Read the transcripts. “Therapy” is healing. Woebot causes or promotes or facilitates “healing.” It is therefore a “therapist.”
There is a lot of merit to talking about why the interaction between therapist and client works. What is it about the “therapeutic exchange” that causes the changes that are desired. It’s hard to tell, really. Outcome measures won’t do it, obviously. Reports of insight won’t do it.
The second question that will arise is what the effect of Woebot is in the long run. Woebot is a phone app. Everybody has a phone. “Therapy,” is therefore available everywhere and everywhen,. How can anyone be opposed to that? That’s what the technologists will have to say.
The therapists will say that what the app will do is the direct the attention of some really needy people away from the humans who can help them. There is really no substitute for the meaningful interaction we can have with other minds, but we can be trained to call something “good enough” even when it is not. If therapy were, as you argue, like scratching an itch, then the client would always know whether the itch has been scratched. But in real therapy, the client realizes whole new options, he constructs whole new approaches to problems. Woebot can’t have that effect on people. What it can do is to teach them that the kind of mechanical “performance of empathy” is all there is and that it is, therefore, “good enough.”
The technologists will have to say that the therapists are denying a lot of people small amounts of real progress because it is not “real” therapy—this despite their failure to define just what “real” therapy is.
Now I think I will go read the article. If they say other things—things I did not foresee that they would have to say—I am going to be really surprised.
 Old fashioned, no disrespect intended, person to person types.
 They focus on the hardware and construct the algorithms that make it do what it does and seem like it seems.
Great analysis. You might want to revisit the posting, there is a lot of duplication.