Their latest effort could have wider ramifications however. They have launched a project to try to understand and predict the suicide rate of veterans. With 22 veterans taking their lives each day, it’s a serious and tragic issue.
The Durkheim Project aims to use machine learning tools to help predict the likelihood of an individual committing suicide in real time.
The system has been collecting data for the past two years in an attempt to train up the predictive models to a level whereby now it is possible to accurately assess risk of suicide 65% of the time.
Central to the data collected is social data. Facebook, Twitter and Google+ are all on board, so the system can analyse what soldiers and veterans are sharing on their social networks and try to gauge their mental health.
The flipside unfortunately is that the project is purely at the research stage at the moment. If the team find an individual they believe is at risk, they are not authorised to act. The hope is however that once the model has proved its worth it can act as the frontline in managing the mental health of both soldiers and veterans, allowing doctors to monitor risk in real time.
Suffice to say that this project is at the moment only looking at the symptoms of the situation rather than the cause. This isn’t just an American issue either of course. It’s believed that 224 vets from the British army killed themselves over a ten year period between 1996 and 2005, which whilst a relatively lower rate than in America, are still far too high.
If DARPA do have a model that is proving useful in predicting risk, hopefully they will share that model with similar organisations around the world. After all, a major concern for UK health officials is a lack of information about veterans living in their area. A lot of the time they simply don’t know who they have on their own doorstep, let alone the risks posed.