In the United States, software helps predict violent police behavior

The technology was studied by a Chicago startup: the software algorithm processes data relating to a single agent and predicts the percentage that a violent act can commit.

(Photo: Roy Rochlin / Getty Images) The decision of the Minneapolis, Minnesota city council to dismantle the police department and replace it with a new public security system has not yet produced concrete results. After the death during a detention of George Floyd and other African Americans and following the numerous demonstrations of the Black Lives Matter movement, in many American cities it has become a priority to find a solution that avoids the onset of other episodes of law enforcement violence to the detriment of citizens, especially those belonging to ethnic minorities.

Among the solutions examined in recent months by the chief of police of Minneapolis, Medaria Arradondo, there has also been the use of software provided by a start Chicago up, Benchmark Analytics, which promises to calculate a cop's chances of committing violent or unfair actions. At the basis of this technology there is an algorithm that processes a series of data - for example, reports from citizens or intervention in particularly traumatic situations such as abuse or murder - this percentage. For the CEO, Ron Huberman, police violence is caused by individual agents, not by the entire department: the software helps to identify and isolate them.

An effective solution?

The news was reported by Bloomberg that, on the merits, it also accepted the opinion of some experts, like Andrew Ferguson, a law professor at American University, who has shown himself to be very sceptical about the effectiveness of this tool. “When you don't trust people, try to trust the data and therefore some companies try to monetize this feeling and sell an immediate solution” . In fact, the software – which is based on the program of the Center for Data Science and Public Policy at the University of Chicago – has not convinced the chief of police of Minneapolis that has not even begun testing. Other cities – approximately 70 according to what refers to Huberman – they have, however, wanted to buy the program and promise to use it within the next year.

The doubts about its actual validity is based on a key factor: the accuracy of the data that are processed by the software. Not all police departments adopt the same methods for the detection of this information. In addition, some surveys have shown that the forces of law and order in the u.s., when it comes to the activation of technological tools, aimed at their recognition (such as the bodycam), often implement voluntarily the practices are wrong . This also adds to the risk, revealed to other tools such as facial recognition , that can be discriminatory for certain individuals, on the basis of race or ethnicity.

The software of the Benchmark, however, is not a novelty of the past few months . Studies and experimentation have been already initiated during the Obama administration, and some cities have been working together for years with the start-up of Chicago. The police of Nashville, for example, is closely connected with the Benchmark from the beginning of 2018, and at the end of July will present a program of prediction of violent behaviour of the cops.