By guest blogger Tomasz Witkowski
“There is but one truly serious philosophical problem and that is suicide” the French author and philosopher Albert Camus stated. But it is not only philosophers who are moved by this issue. Psychologists are seeking ways of preventing this tragic death, and health care organisations are sounding the alarm. Around a million people die at their own hand every year, which makes suicide the tenth most common cause of death. Additionally, for every completed suicide, there are 10 to 40 survived attempts, which means that in the USA alone 650,000 people each year are taken to emergency rooms following an attempt on their own life. Yet what is most disturbing is that the number of suicides is continually rising. The WHO reports that since the 1960s this number has grown over 60 per cent.
Is psychology capable of identifying the risk factors that can push people to take their own lives? Joseph Franklin at Florida State University and his research team at the Technology and Psychopathy (TAP) Lab have provided an answer, but it is a disappointing one. Our capacity to predict whether someone will make a suicide attempt is no better than chance. What is worse, we have not made any progress in this area in the last half-century. These striking conclusions come as the result of a meta-analysis of 365 studies into suicide risk conducted over the last 50 years and published recently in Psychological Bulletin (pdf).
Should these results be seen as discrediting psychology, leading us to declare its incompetence in that sphere and that the matter should be left in the hands of philosophers? I would not be so quick to make that argument, although I do consider the findings to discredit a certain manner of practicing psychology. As Franklin and his colleagues observed, the problem with past research is that the methods have been extremely narrow. Most studies only looked at a single risk factor, such as depression or low serotonin in the brain, and then followed patients over a decade, while in real life it is generally the interaction of many factors of a given impact level that provide the fuel powering some sort of dramatic reaction.
Moreover, most often suicide is not caused by the simple sum of those factors, but from a complex interaction between them, yet the research methods generally applied by psychologists are not capable of uncovering such dynamics. Research on complex systems, which is most certainly an appropriate designation for the mind, long ago demonstrated that their characteristics are not always discernible when looking at constituent elements. Why, then, do the vast majority of psychologists so obstinately insist on treating people as simple machines that react linearly in response to simple stimuli?
I know of only one answer to this question. The single-factor approach is easier and generates the effects expected by scientists, such as grants for research, the conviction that they are engaged in important work, points for publications indispensable in making progress in one’s scientific career, participation in conference tourism, etc. Meanwhile, machines are increasingly overtaking researchers trapped in this mode of thinking. In the past two years, multiple groups have begun work developing “machine learning algorithms” to combine tens or even hundreds of risk factors together to predict suicidal behaviours. “The preliminary results are promising, with algorithms predicting suicidal behaviours with greater than 80 per cent accuracy, but this work is just in its initial phase,” said Franklin. “However, in the very near future, this work may produce accurate predictions of suicidal behaviours on a large scale.”
His team has already developed a free web app that has proven effective in trials at reducing suicidal behaviours. The app, called “Tec-Tec” uses a simple game to train users to see certain words and images in a different light (a form of “evaluative conditioning”) and is available on iTunes and Amazon right now. Studies have shown that the app reduced suicidal behaviours by about 50 per cent over the course of a month in hundreds of people, and they hope in the future to reach a reduction rate approaching 100 percent.
Does this mean that all the existing scales measuring the risk of suicidal behaviours should be thrown away, and the counsellors and therapists working on suicide prevention should think about getting a new job? In fact Franklin recommended therapists continue using the guidelines, but he added that there’s an urgent need to re-evaluate them.
Franklin’s cautious approach to the tools currently in use arises primarily from the fact that past studies have not looked at the impact of known risk factors over short time-spans (on average, they examined one risk factor over 10 years) where they may be more useful. Furthermore, previous studies examined only a portion of the measures from the far more expansive set that is used in clinical assessment.
Franklin also is likely aware that an experienced counsellor frequently relies on her or his own experience and clinical intuition together with more formal risk measures, and that these clinical judgments and measures together might be more effective than a mere coin-flip. Unfortunately, personal clinical intuition does not constitute a cohesive and reliable system of knowledge with significant predictive capacity: it can’t be taught or used by others. It would appear that to prevent more suicides in future, artificial intelligence could be a far better solution.
Post written by Dr Tomasz Witkowski for the BPS Research Digest. Tomasz is a psychologist and science writer who specializes in debunking pseudoscience in the field of psychology, psychotherapy and diagnosis. He has published over a dozen books, dozens of scientific papers and over 100 popular articles (some of them in Skeptical Inquirer). In 2016 his latest book Psychology Led Astray: Cargo Cult in Science and Therapy was published by BrownWalker Press. He blogs at https://forbiddenpsychology.wordpress.com/.
Read an interview with Tomasz in the latest issue of The Psychologist.