Every year close to 800,000 people commit suicide which is one person every 40 seconds, according to the World Health Organization. It is a global problem and has become the second leading cause of death among those aged 15-29 years old. And for every suicide, there could be at least 100 to 200 attempts.
While those figures are bad enough, things look even worse when you consider that half a century of suicide prevention research has not produced any real progress. That was the conclusion of a meta-analysis on the topic by FSU Psychology Assistant Professor Joseph Franklin. His findings are stark. They indicate that while traditional risk factors such as substance abuse and depression can predict suicidal thoughts and behaviours they can only do so with an accuracy that is marginally better than random guessing.
Every suicide is a tragedy and many cases are preventable. Prevention is a complex issue that involves lots of different factors, but spotting warning signals and taking them seriously can be key. While there may be clear signs that someone is thinking about taking their own life this is not always the case. Inner pain isn’t obvious. If it was, suicide wouldn’t be the public health problem it is today.
AI and Suicide Prevention
But the situation could one day improve thanks to technology. Artificial intelligence which is being hailed as a modern equivalent of the industrial revolution may be able to do better than humans when it comes to gauging suicidal intentions, identifying people earlier and more accurately.
Studies have already shown that it’s possible. In fact, AI can accurately predict suicide attempts up to two years in advance. A 2017 study led by Florida State University researcher Jessica Ribeiro revealed that machine learning algorithms and artificial intelligence could predict with 80 to 90 percent accuracy whether or not someone will attempt to take their own life. The accuracy is even greater as a person’s attempt comes closer. For example, one week before an attempt is made the accuracy is 92%.
Her paper, tilted Predicting Risk of Suicide Attempts over Time through Machine Learning was published by the journal Clinical Psychological Science. Using anonymized electronic health records from two million patients in Tennessee, algorithms assessed a number of factors such as pain medication, prescriptions and the number of ER visits to learn which combination of factors increased the likelihood of someone trying to commit suicide.
“This study provides evidence that we can predict suicide attempts accurately,” said Ribeiro. “We can predict them accurately over time, but we’re best at predicting them closer to the event. We also know, based on this study, that risk factors — how they work and how important they are — also change over time.”
Another institution at the forefront of this burgeoning and fascinating field is the Center for Artificial Intelligence in Society (CAIS) at the University of Southern California, according to a recent blog post from the MSW@USC. Its researchers are in the beginning stages of suicide prevention projects. They plan to use artificial intelligence to dynamically model social networks and the messages transmitted across them. This involves collecting a lot of data about how social networks and what people think about their networks affect their mental health. One of the purposes of this is to identify key targets for intervention. Currently, CAIS is working with three different populations: college students, homeless youths and the military.
Among the military, the risk of suicide is greatest during periods of transition, such as on joining up and when they first return from deployment. Therefore CAIS researchers plan to interview members of the military before deployment and upon return to see if they can identify a pattern of merging suicidal thoughts.
With regards to college students, researchers will look at friendship networks to figure out how to select students who will be able to look for signs of depression and suicidal thinking among their peers. Algorithms will perform sophisticated social analysis looking for students who have the most influence on others.
To help prevent suicide among the homeless researchers will reverse engineer their social networks to see what would work best and where best to insert assessment tools to provide help. This approach will also be used with active duty service members.
Others that are looking to artificial intelligence to stop people taking their own lives include the Department of Veteran Affairs. Tragically, an estimated 20 veterans a day take their own life. Researchers are aiming to create an AI-bolstered assessment program that uses patient-specific algorithms to identify behaviours that could indicate suicidal inclinations.
Facebook’s AI for Suicide Prevention
Already the technology is being used to help detect suicidal thoughts and feelings.
In November 2017, Facebook rolled out its artificial intelligence suicide prevention tool. The social network giant has developed pattern recognition algorithms that spot warning signs in live feeds, users’ posts and the comments their friends leave in response. They were trained on posts that had previously been flagged.
Although the tool is available in several territories it has hit a snag in Europe and cannot be deployed. That’s because data protection laws ban the processing of an individual’s sensitive personal information without their consent.
The field may well be new with a lot of research hours lying ahead, but so far the signs are promising that artificial intelligence could one day be used as a powerful predictive tool that will allow health workers to intervene before someone attempts to take their own life.
Right now, the only way anyone knows if someone is having suicidal thoughts is if they decide to talk about them. Given that many individuals who attempt suicide do not admit it beforehand, the need for a radical new approach such as that offered by artificial intelligence is urgent.