It’s surely law enforcement’s ultimate aim: to predict crimes before they happen. IBM has been working on using its systems to analyze patterns of criminal behavior, then using this data to predict where future crime may take place. But a report in the journal Psychological Medicine is attempting to take this one step further: a team of researchers has developed a complex algorithm that identifies U.S. Army soldiers most likely to commit crimes of a violent nature.
If this sounds like something straight from the movie Minority Report, then you wouldn’t be too far off the mark. Although the movie’s PreCrime organization used the abilities of clairvoyants to detect criminal actions – particularly murder – the military equivalent uses a series of algorithms and a vast amount of background information to make its predictions. Also unlike the Philip K. Dick-inspired movie, the potential criminals will not be arrested before the crime is due to take place, but rather given counseling aimed at curtailing any urges to commit violent acts.
Using the military records of almost 1 million soldiers who served in the U.S. Army over a six-year period between 2004 and 2009, the team of researchers focused on those that committed violent crimes, including murder, manslaughter, kidnapping and armed robbery. Domestic violence and sexual assaults were not included in this study as these are thought to be influenced by other, more complex factors.
Hundreds of different variables for each soldier were considered, including their demographic, health history, economic history, history of any behavioral problems, place of residence, history of criminality, and army career.
The highest risk group was found to be young, poor soldiers with a low rank, a history of disciplinary trouble and attempted suicide, and who have been recently demoted. The highest risk male group, accounting for 5% of the total male population, was responsible for 36% of the crimes perpetrated by men; the 5% of female soldiers at highest risk accounted for 33% of female crimes, although females were half as likely overall to commit violent crimes as their male counterparts.
A predictive algorithm based on these results was then constructed and subsequently tested on a sample of more than 43,000 soldiers that served in active military units between 2011 and 2013. They found that those categorized as belonging to the 5% highest risk group committed 51% of the violent crimes during that time period.
Mental health is an enormous issue in the U.S. Army, with the Los Angeles Times reporting that nearly one in five soldiers had a common mental illness, including depression or panic disorders before enlisting in the army.
Slate references a 126-page U.S. Army report released in 2009 that found the murder rate around the Army’s third-largest post, Fort Carson, doubled since the start of the Iraq War six years earlier. A unit of soldiers – the 2nd Battalion, 12th Infantry – spent a large part of their 15-month tour in Dora, one of the most dangerous parts of Baghdad. It is probably no coincidence that the unit’s post-traumatic stress disorder (PTSD) diagnosis rate was more than 300% higher than a like-for-like unit that spent its tour in a far less violent part of the war-torn country.
Although one definitely cannot state that having combat PTSD automatically results in violent crime, several studies do describe a link between combat PTSD and the likelihood of crimes being committed after serving in an active military unit.
This new predictive system appears to be an efficient tool for the U.S. Army to identify high-risk soldiers without waiting for the crimes to take place first. But what they do with this information will require careful thought and ethical consideration, especially since a large proportion of those categorized in the high risk group do not go on to commit crimes. To some, however, this controversial program will be seen as a state-funded breach of personal freedom.
[H/T: LA Times]