What does it look like to you? Lambert / Getty Images
What does this Rorschach inkblot, a popular psychological exam given to psychiatric patients just up until a few years ago, look like to you? A sneering clown with a cone hat? A Jewish Star of David? How about the impending takeover by automation and robots of the entire field of psychiatry?
Yup, that one.
This isn't just speculative, either. A recent study found that a computer analyzing speech patterns in a pool of patients was better at predicting the development of psychosis down the line than clinical practitioners were. In fact, the computer program predicted those who would suffer a psychotic episode (and conversely those who wouldn't) with 100 percent accuracy. There literally is no getting better than that.
Here's how the experiment "” performed by teams from Columbia University Medical Center, New York State Psychiatric Institute, and the IBM J. T. Watson Research Center "” went down. A computer program monitored how a pool of subjects answered questions, constructed sentences, drew out, cut short, and connected thoughts, exaggerated or withheld reasoning, etc. The human researchers identified speech as an important indicator of possible psychosis, and so it was the focal point in this study. The machines analyzed the answers for content as much as form, and then proceeded to make estimates as to the future mental health of the subjects.
Yes, but can a robot see its parents making love? RDA / Getty Images
What did they find? Startling results in fact. Turns out that of the 34 individuals, five suffered a psychotic episode two-and-a-half years after the initial exam, and 29 didn't. The computer program guessed the fate of each one...perfectly. What's more is the technology is cheap to employ, returns answers quickly and doesn't require intensive, invasive procedures. The researchers recognize that the same experiment will need to be repeated on a larger scale to test for repeatability and reliability, but if it turns out that these physicians assistants can help diagnose and prognosticate, it could be a real game changer for the field of mental health.
Because the automation revolution is no longer the stuff of science fiction anymore. This past May, National Public Radio reimagined some of the findings in Frey and Osborne's 2013 paper "The Future of Employment: How Susceptible Are Jobs to Computerization," filing different jobs under, essentially, one of two categories: at risk of automation, and not at risk. In reality their measurements were reported on a sliding probability, but let's make things easier on ourselves, because they're about to get pretty tough.
On what criteria did these scientists assess automatability? Well, let's compare two different jobs, one at a high risk and one at a low risk.
Bookkeeping, the forgotten art of sliding numbers into their appropriate columns in a spreadsheet and pressing Enter repeatedly until you get the output that doesn't make your ulcer palpate, is predicted to be totally demolished under the robotic stampede. NPR marked it at 97.5 percent likely that it would be automated in the near future. It's little more than dragging and dropping, and can be easily made obsolete by a simple algorithm.
On the other hand, clinical counseling and school psychologists have a remarkably low chance of being automated, at half-of-one percent in the near future. Why is this? Because clinicians must be able to pull clever solutions out of their bums, assist on a personal (see root: person) level, they aren't performing manual labor that demands they do weird things with their bodies, and they have to negotiate. These are all attributes of a hard-to-automize job, and which all favor the human worker.
At the moment, and for the foreseeable future, there certainly will be a need for human clinicians, especially considering how green this experiment is and how it still would need to pass through several rounds of red tape before it can be implemented diagnostically by physicians.
Sitting pretty...for now Education Images / Getty Images
But I'd argue that diagnosis is a huge percentage of what a psychiatrist actually does. The role of a psychiatrist and a psychologist differs, in addition to others, in one major area: the prescribing of medication to treat a disorder. Once a psychiatrist diagnosis a patient, he or she proceeds in prescribing a drug or cocktail of drugs to treat the disorder, monitoring progress or setbacks, adjusting and then readjusting dosage or even the type of med.
The psychiatric approach varies greatly from the psychological one: the former's focus is on medication, the latter on taking behavioral and cognitive and other approaches to improving mental health. The latter is, arguably, more solution-based and personal, and involve higher degrees of negotiation and less manual labor (like filling out scripts). Now, compare this dichotomy to another trend, that of the skyrocketing use of anti-depressant medications in the United States.
Anti-depressant meds, to reiterate, can only be described by psychiatrists, and what an excellent job they're doing at it, too (note the sarcasm). Today, one in 10 Americans is on anti-depressant medication, and for women in their 40s and 50s that ratio drops to one in four. As depression and anxiety disorders become more ubiquitous, it's not unlikely that we will see the pill prescription numbers to follow suit.
To pill or not to pill... BSIP / Getty Images
But what does the rise in depression and other "pharmaceutically treatable" mental disorders have anything to do with robots? Well, we've already seen that computers can help, flawlessly, in predicting future disorder. That's a major first hurdle in the rehabilitation process. What's to say that they won't be employed at every stage of therapy thereafter? For Pete's sake, the study this whole article is centered around was partly commissioned by the New York State Psychiatric Institute: psychiatric is in the name!
If computer programs can more accurately predict and diagnose psychosis, and they're used in developing the drugs later used in treatment, what's to say that they won't be used in the monitoring and feedback checkpoints as well? Psychiatry, unlike psychology, is a numbers game, except instead of odd, even, black and red, the numbers correspond to neurochemicals, and what's at stake is the well-being of the mind. How many milligrams of Zoloft need to be prescribed so that the patient feels joy again, but not so much that they rebound in the opposite direction? A delicate numbers gain, that's all.
And we've already seen robots dominate in fields where numbers are concerned: the stock market, chess matches, theoretical mathematics, to name a few. Is psychiatry the next, numeral frontier for our Robot MDs? I'd bank on it.