Investigation Finds 'Predictive Policing' is Wildly Inaccurate

Using Artificial Intelligence to predict crime has grown in popularity with law enforcement agencies around the world, but there is little evidence that the technology is effective in reducing crime or costs.

An investigation has shown that so-called predictive policing software produced almost no positive results, adding to a previous study that showed such software disproportionately focused on low-income and minority neighborhoods.

Predictive policing software has been growing in popularity with police forces around the nation. The companies that sell their services to police departments claim they use large data sets to emotionlessly and without bias predict where crimes are most likely to occur.

But the investigation, performed by two tech-focused media outlets, revealed that at least in one locale, the software provided hardly any accurate predictions. The investigation looked at 23,631 predictions made by software from a company called Geolitica (previously PredPol) in the city of Plainfield, New Jersey.

It showed that over 10 months, Geolitica software correctly predicted where crime would occur less than one-half of 1% of the time.

David Guarino, a captain for the Plainfield Police Department said the city rarely used the program and has since abandoned it.

“Why did we get PredPol? I guess we wanted to be more effective when it came to reducing crime. And having a prediction where we should be would help us to do that,” Guarino said. “I don’t believe we really used it that often, if at all. That’s why we ended up getting rid of it.”

It has been reported that parts of Geolitica are being purchased by SoundThinking, which provides technology services, including predictive policing, to law enforcement.

Despite their inefficiencies, there is no guarantee their growth in popularity will be stemmed. In 1987, the first speed camera was installed in the United States. Today, there are more than 5,000 speed cameras installed across 18 US states and the District of Columbia, despite a myriad of controversies about their accuracy.

The city of Chicago reportedly gave out $2.4 million in erroneous fines issued by speed cameras between 2013 and 2015.

A report in 2014 revealed that Washington, DC, speed camera reviewers would make arbitrary decisions about which car to fine when more than one was in the picture.

In Baltimore, Maryland, an internal audit in 2013 revealed that cameras, promoted as having less than a 2% inaccuracy rate, actually had error rates exceeding 10%. One camera was found to be inaccurate 58% of the time. The audit was hidden from the public for nearly a year before complaints from residents that their cars were being ticketed while parked sparked a media investigation.

Despite this, speed cameras are still in use in the aforementioned cities and have continued to increase nationwide. This, critics suggest, is because the programs are big money-makers for cities and law enforcement.

Predictive policing does not have as direct of a link to police department revenue as speed cameras, but its promotional materials claim it can save departments money and a study by the RAND Corporation stated that the software is generally perceived as beneficial by officers, despite that study also showing no statistically significant reduction in crime in areas it was used.