Could a Machine Identify Suicidal Thoughts?

A Netflix series titled, "13 Reasons Why" unravels the reasons why a teenage girl decided to commit suicide. The subject of suicide and suicide prevention in society has gone from too sensitive to discuss, to a worldwide effort to discuss suicide in order to prevent as many victims as possible. I personally have lost two close friends to suicide. So I, along with Dr. Marcel Just ( a cognitive neuroscientist at Carnegie Mellon University) and Dr. Matthew Nock ( a clinical psychologist at Harvard University), are committed to finding ways to recognize those who experience suicidal thoughts, and those who have attempted suicide before.

The background information previous to their study and experiment declared suicide as the second-leading cause of death in middle-aged Americans, and that those rates are only increasing. The study of the brain is the source of information of the body. This is where both Dr. Just and Dr. Nock conducted their research. They wanted to find a reliable biological predictor of suicidal thoughts and actions. Their study combined neural imaging with machine learning to discover if and how the brain responds to positive an negative words relating to life and death. Nock previously used implicit association test in order to predict those with suicidal thoughts or previous suicide attempts. This technique had the ability to correctly identify 9 out of 17 suicidal subject. Just conducted an experiment using machine learning. Impressively this test resulted in a 80 to 90 percent accuracy rate in detecting suicide risk. With this method of experiment we can see the actual thoughts people have about suicide

Both of these studies were combined in order to create an even more accurate method. Nock contributed by his accurate ability to detect suicidal thoughts through his implicit association tests. Just contributed with his technique of functional magnetic resonance imaging, or fMRI. This allowed for the use of both methods to create a scanner. Subjects were placed in the scanner and asked to think of words to describe suicide. The scanner recognized thoughts like: death, trouble, carefree, cruelty, praise, and good. Then was asked to determine whether or not the victim was suicidal or not. This method proved to be 91% accurate. In phase two of their study the fMRI exposed the brains response to certain words grouping the responses within 4 categories: anger, shame, sadness, and pride. The scanner was trained to distinguish between the brain activity that was suicidal and the brain activity that was not in correlation to the words expressed compared to the brain's activity.  The scanner accomplished the task with 91 percent accuracy, identifying 15 out of 17 suicidal subjects and 16 out of 17 controls (non suicidal subjects) correctly.

Other scientist have studied the results of Nock and Just and find their method to be, although accurate, being preliminary and expensive. Nock and Just agree to the fact that their study is not concluded. Their goal is to find a method that is both extremely accurate, and more easily accessible and available to the common folk. They are working towards finding new pieces to the puzzle of suicide through their research.

Work Cited

https://www.scientificamerican.com/article/could-a-machine-identify-suicidal-thoughts/