I Challenge Watson To A Game Of Therapy!

March 4, 2011
H. Steven Moffic, MD

Watson, as many will know, is not necessarily the name of a human, though the Watson of DNA fame may come to mind. Rather, in this context I am referring to IBM's artificial intelligence computer system.

"What a piece of work is man." -  Hamlet

Watson, as many will know, is not necessarily the name of a human, though the Watson of DNA fame may come to mind. Rather, in this context I am referring to IBM's artificial intelligence computer system. It can answer questions in a natural language. To test the extent of this ability, it recently competed in 2 quiz games of Jeopardy! against 2 human record holders in this game. Watson won convincingly both times.

This was viewed as an advancement of the 1997 victory of the Deep Blue computer over Gary Kasparov in chess. Jeopardy! was considered a much more complex challenge than chess since it included words.

Besides its computational skills, Watson seems immune to psychological tactics in games, such as bluffing and intimidation. Now, wouldn't that skill be a help for my work in prison!

Some are suggesting that next up for Watson and the like are more important challenges, such as aiding in the diagnosis and treatment of patients. And, why not--at least for general medicine and surgery? Already, among other things, computers and the internet help in anticipating dangerous medication interactions. Electronic medical records can provide some prompters for guiding the clinician. Computer interaction with patients can elicit basic and ancillary information. Robots are revolutionizing surgery. Physicians are starting to use social networking.

Computers can now even mimic human emotion. Robots can be programmed to respond in some appropriate ways to certain stimuli and situations. Of course, the robots can't feel the emotions, can they?

In ever-expanding ways, computers seem to be augmenting some human capabilities. Raymond Kurzweil, a leader of the Singularity movement, predicts that the exponential development of computers will surpass the brainpower of humans in a decade and eventually our brains could become immortal in a "transhuman" merger with computers. Eric Schmidt of Google fame predicts that computers will eventually be able to guess what we are thinking. Now, I thought that was what psychiatrists and psychologists did best, or at least the public thought so. As a bonus, presumably computers wouldn't have the boundary temptations we humans do.

So, it must then seem grandiose, perhaps to a psychotic degree, for me to challenge Watson. Given my own pathetic computer skills, I'm even probably vastly underestimating what Watson could do. Couldn't Watson much more quickly and accurately put together signs, symtpoms, and historical information to make a cookbook DSM diagnosis? Couldn't it then incorporate this information into available expert guidelines on psychotherapy and medication algorithms? Watson should know when polypharmacy was really necessary and not a desperate attempt to help. Why, maybe the APA should even consult with Watson for the final version of the upcoming DSM-5!

But aren't there things that Watson and computers can't do so well that may be especially important for psychiatry and psychotherapy? So far, and maybe forever, even with the reflexive emotional reactions that can be programmed into robots, they can't replace our complicated--and still not well understood--emotional intelligence. This kind of intelligence includes being potentially able to detect and decipher emotions in others, as well as in artifacts and ourselves; the ability to understand the causes and consequences of emotion; the ability to harness those emotions to make decisions; and the ability to manage emotions in oneself and others. Emotions are also the link between moral reasoning and moral action.

Perhaps this is why the hoped for computer therapists has not nearly reached its early promise. Way back in 1964, Eliza was modeled after the famous psychotherapist Carl Rogers, and used reflective language that seemed non-directive and supportive. Many at first thought they were having a therapeutic experience. Enthusiasm ran so high that reputable scientific commentators enthused that hundreds of patients an hour could be treated in this way, very cheaply, and perhaps on so-called computer psychotherapeutic terminals. The originator of Eliza, Joseph Weizenbaum of MIT, actually became concerned about this and pulled the plug on the Eliza project. Fortunately, this was before managed care could get hold of it!

More recently, what has evolved more slowly was discussed in a 2008 article by Dr. Wright in our Psychiatric Times. He describes how the use of computers has become more refined in the form of computer-assisted therapy (CAT). This includes virtual reality exposure therapy and self-help computer programs. Relevantly, research now indicates that the involvement of a human therapist significantly increases the completion and usefulness of CAT.

Another kind of computer versus humans test was reported on in the article Mind vs Machine in the March, 2011 Atlantic. In 2009, in the annual Turing Test to find out whether a computer can act "more human" than a person, the humans (including the author of the article) won convincingly. Computers were rigid and inflexible compared to the more sensitive and responsive human testers. Though skilled at logical analysis, computers fared much less well at adaptive goal-setting. Like HAL in the movie 2001, they did not seem so able to resolve psychological conflicts very well. What humans seemed to do best in the test may be akin to what therapists do best in therapy.

Now maybe I wouldn't bet against me, or you. The ability to be empathetic and convey warmth, positive regard, and compassion are still the strongest predictors of a postive therapeutic outcome. These abilities, if we don't lose them, are what may make us more essential over the long run than other medical specialists. Couple this with high emotional intelligence, wisdom gained from experience, and sound therapeutic skilles, and we might be surprised at how well we could do against Watson in a game of Therapy!

Watson, if you read this, I'm ready!