There’s a recurring theme in David Fincher’s brilliant 2007 movie Zodiac concerning the various ways potentially crucial pieces of crime scene evidence fall through the cracks as the result of poor information-sharing between police departments.
More than 40 years after that movie was set, things have moved on in a big way, but there’s still the problem of data silos — meaning isolated pockets of data that aren’t shared and cross-referenced in the way they should be. That’s a massive problem because, as William Wong, a computer science professor at the United Kingdom’s Middlesex University London explains, criminals aren’t always easily categorized.
“If I’m a criminal, I might usually be a house burglar, but if I see a car that’s unlocked I may burgle it, even though that’s not what I usually do,” Wong told Digital Trends. “Because of this, it’s important that investigators can search across different data silos.”
Middlesex University is one of a number of international universities working on a new system called VALCRI, which is designed to solve this challenge, among others. Short for “Visual Analytics for sense-making in Criminal Intelligence analysis,” VALCRI is an automated Sherlock Holmes-style crime-solving computer system that uses the latest machine learning tech to scan through masses of police records, relevant interviews, crime scene photos, videos, and a whole lot more, and then find links where they might not be obvious. The project began back in 2014, when the University of Middlesex benefited from a $17 million investment to kick-start research. Since then it has gone from strength to strength.
“What we’re doing is creating machine augmentation tools that allow people to analyze crime data in new dynamic ways,” Wong said. “It’s about joining the dots. The problem in a lot of cases is that investigators don’t know what the dots are that they need to connect.”
VALCRI focuses on breaking down classic fixed crime categories, and instead turning criminal profiles into sets of behaviors that can be easily searched. In this way, Wong’s proverbial car thief opportunist might have his or her profile flagged for the crime, even if they’ve only ever previously robbed houses. It’s also able to suggest lines of research that investigators may want to follow — and present all of this data in a highly visual touchscreen interface that makes every link or piece of evidence easy to analyze in detail.
Rather than just presenting more information, the ultimate idea is to use its smart AI tech to come up with hypotheses about the “how” and “why” a crime has been committed, en route to finding the “who” involved.
As of now, VALCRI is still being tested in the U.K., but Wong and his colleagues are confident that it could represent the future of policing. Minority Report, here we come!
