Description: Suicides are preventable, but, ….. predicting suicide attempts so that timely assistance and support can be offered is incredibly difficult. Part of the problem, from an epidemiological perspective, is that suicide is a statistically rare event (annually 11 per 100,000 per year in Canada currently) but still, 10 people a day commit suicide in Canada. Another part of the problem, related to suicide’s statistical rarity, is that there is no single or small number of factors or variables or signs or symptoms that predict both the possibility and the timing of suicide attempts. The number of potentially contributory factors is huge and as such very hard to gather, monitor through time, and, for the clinicians doing that work, very hard for even well-trained clinicians to get their heads around and offer help to people at the right times. What to do? Well, hmmm, big complex data and a need to make individually relevant predictions… how about writing an algorithm and having a machine do the job? Have a read through the article linked below to see where this possibility is current at and to see some of the ethical and technical challenges currently at play. Oh and see what one such algorithm had to say about Anthony Bourdain’s tweet stream over time and in the days leading up to his suicide.
Source: Can an algorithm stop suicides by spotting signs of despair? Erin Anderssen, The Globe and Mail.
Date: November 24, 2018
Image Credit: Chris Pizzello/The Associated Press
One of the core challenges of clinical psychology and of working in areas of mental health is that much of the data needed to connect and work with people is inside or psychological and thus hard to access, especially if the person in question is socially withdrawn and considering suicide. Simply making contact and then deciding who needs assistance NOW is very difficult. Many risk factors can be at play and most are actuarial, meaning that they simply say that relative to the general population a subset of people is more likely to think or do X, Y, or Z. And, in the case of low frequency actions like suicide, many of those in the “at-risk” sub-group will NOT take the feared action. Looking at tweets and other forms of social media potentially provides us with, over time, glimpses into people’s thoughts and moods and as such processes like Dr. Kaminsky’s algorithm might provide us with opportunities to offer timely assistance to those with suicidal thoughts. Of course, there are ethical issues related to privacy and related to the still present and strong stigmas associated with mental illness and thoughts of self-harm that will need to be considered. Computer algorithms cannot relieve us of those ethical responsibilities. However, possibility of reaching those with suicidal thoughts and plans and reducing their suffering and that of their friends and relatives is a powerful motivating force.
Questions for Discussion:
- Why is suicide so hard to predict?
- What are some of the ethical issues that arise from sing an algorithm to monitor twitter streams or of having Facebook search for word phrases associated with psychological pain, illness, or despair?
- Health Canada is funding research into how an algorithm approach might work in Canada. What might/should some of the guiding parameters of that work be?
References (Read Further):
Franklin, J. C., Ribeiro, J. D., Fox, K. R., Bentley, K. H., Kleiman, E. M., Huang, X., … & Nock, M. K. (2017). Risk factors for suicidal thoughts and behaviors: A meta-analysis of 50 years of research. Psychological Bulletin, 143(2), 187. http://psytaplab.com/s/Franklin-et-al-2016-overall-meta-of-STB-risk-factors.pdf
Guintivano, J., Brown, T., Newcomer, A., Jones, M., Cox, O., Maher, B. S., … & Kaminsky, Z. A. (2014). Identification and replication of a combined epigenetic and genetic biomarker predicting suicide and suicidal behaviors. American journal of psychiatry, 171(12), 1287-1296. https://ajp.psychiatryonline.org/doi/pdfplus/10.1176/appi.ajp.2014.14010008
Walsh, C. G., Ribeiro, J. D., & Franklin, J. C. (2017). Predicting risk of suicide attempts over time through machine learning. Clinical Psychological Science, 5(3), 457-469. http://psytaplab.com/s/Walsh-Ribeiro-Franklin-proof-version-ML-and-sui-attempt-prediction.pdf
The Globe and Mail (August 30, 2018) Federal health agency to mine social media for study on suicide trends, risk factors, https://www.theglobeandmail.com/life/article-federal-health-agency-to-mine-social-media-for-study-on-suicide-trends/ accessed November 25, 2018.