We all know that social media is the criminal's worst enemy. But this summer, a group of MIT researchers are collaborating with police to test software that can reliably predict whether a person is part of a gang based on their social networks, building on similar software used to track insurgents in Afghanistan.
For the past decade or so, the Army has used a number of similar pieces of software to visualise the relationships between suspected and known insurgents involved with the development of improvised explosive devices. It works by grouping many individuals by levels of association — and by doing so, gives researchers a way to recognise system-wide patterns and predict who might be involved in a particular operation.
According to MIT's Technology Review, there are plenty of sociological similarities between insurgents and American street gangs. "In the last 10 years or so, researchers have revolutionised the way military analysts think about insurgency and the groups of people involved in it," explains the Review. "Their key insight is that insurgency tends to run in families and in social networks that are held together by common beliefs."
So it makes sense that the insights gleaned by Army intelligence could help out police officers at home. That hunch was confirmed on June 28th, when a group of MIT researchers published a paper in Physics and Society that details how a similar software is being used to track gang violence.
The software is called Organisational, Relationship, and Contact Analyzer, or ORCA, and it groups people in a particular community by their known relationships, as well as their arrest records. Based on the algorithm, they can predict whether a particular person is likely to be a gang member; It's also able to map "corner crews," which operate hyper-locally, and "seed sets," or individuals who are highly influential.
For this particular study, ORCA was tested on a three-year history of 5400 arrests. Based on those numbers, it revealed 11,000 relationships, created a network map of 468 members belonging to about 20 gangs. The analysis is continuing through this summer in a "major metropolitan area," though they won't name which one. Eventually, the software could become ubiquitous in police stations nation-wide. It seems that crime, just like business, all comes down to relationships. [Technology Review]