Numerous law enforcement agencies in the United States embarked on an ambitious journey to employ AI-based software to predict crimes, reminiscent of the futuristic concepts depicted in the movie Minority Report. However, the outcome turned out to be far from the envisioned success. The AI program’s performance was abysmal, with a mere 0.5 percent success rate, indicating that it was unable to accurately forecast criminal activities before they occurred.
This low success rate suggests that the AI software failed in its primary objective of preemptively identifying criminal behavior. In the world of Minority Report, a much higher level of accuracy was portrayed, but in reality, the complexities of predicting human behavior and preventing crimes proved to be much more challenging for the AI system.
The experiment’s outcome underscores the significant limitations of current AI technology when it comes to predicting future criminal actions. While AI has shown promise in various fields, its application in the realm of crime prevention remains a formidable challenge that requires further refinement and development.
For more than a decade, law enforcement agencies in the United States and the United Kingdom have been using “predictive policing” software. This software relies on algorithms and artificial intelligence similar to what scientists use for predicting earthquake aftershocks. Their aim has been to predict and prevent crimes before they occur, much like the futuristic scenario in the popular movie Minority Report.
However, a joint investigation conducted by The Markup and Wired has revealed a troubling reality. The approach, which was expected to be highly successful, has fallen far short of its intended goals.
One illustrative case is the Geolitica predictive policing software, used by the Plainfield, New Jersey police department. This department was the only one out of 38 that provided data for the investigation. The results were remarkably disappointing. Geolitica’s ability to predict crimes proved to be highly inaccurate, with a success rate of less than 0.5 percent. In other words, it was not effective at all in anticipating and preventing criminal activities.
To reach their definitive conclusion, The Markup and Wired undertook a comprehensive examination of the predictions generated by Geolitica. This analysis encompassed a staggering 23,631 forecasts made by the software during a period spanning from February to December 2018. The findings of this investigation were nothing short of astonishing, as they unveiled that fewer than 100 of these predictions actually aligned with real instances of criminal activity. This shockingly low number translated to a success rate of less than half a percent.
It’s worth noting that Geolitica’s algorithm displayed marginally better predictive accuracy for specific categories of crimes. For instance, it managed to correctly identify 0.6 percent of robberies or aggravated assaults, in contrast to the 0.1 percent accuracy it achieved in predicting burglaries. Nevertheless, these nuances did not significantly improve the overall assessment of the software’s effectiveness.
In light of Geolitica’s underwhelming performance, Captain David Guarino of the Plainfield Police Department, which was one of the agencies utilizing the software, candidly shared his perspective. He questioned the rationale behind their adoption of Geolitica, stating, “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. I don’t know that it did that.” Guarino further revealed that the software was seldom, if ever, put to practical use, leading to the ultimate decision to discontinue its usage.
In addition to concerns about the software’s effectiveness, Guarino also raised questions about the financial investment it entailed. The initial subscription fee amounted to $20,500 annually, with an extra $15,500 for a second year of usage. He suggested that these funds might have been better directed towards supporting community programs or other initiatives aimed at addressing the underlying causes of crime.
Furthermore, Wired reported that Geolitica is on track to cease its operations by the end of the year. However, the company’s personnel have already found a new home within SoundThinking, formerly known as ShotSpotter, another player in the law enforcement software industry. As a result, customers who had been relying on the soon-to-be-discontinued predictive policing platform are expected to transition their software needs to SoundThinking, ensuring a continuity of service.