There is a simple question to get into the matter: Would you rob a bank? Would you kidnap a stranger for ransom? Would you pick the lock of a van parked three blocks from your house and then resell it? It doesn’t matter if you answered yes or no. A group of scientists are working at the University of Chicago convinced that the answer lies in some way with an algorithm, a program capable of anticipating crimes with amazing accuracy.
The tool is “fed” public data on violent or property crime. With this raw material, he identifies temporal and geographic patterns that help him predict certain crimes. And it does not do anything wrong, at least according to the data handled by its creators: they ensure that it can anticipate them a week in advance and with 90% accuracy.
To shake off prejudices or possible biases, the model uses grids instead of the traditional concepts of neighborhood or administrative boundaries. The algorithm divides the city into mosaics about a thousand feet wide – just over 300 meters – for which it predicts crime levels. The researchers tested it with data from Chicago, but also from Atlanta, Austin, Detroit, Los Angeles, Philadelphia, Portland and San Francisco. In all, they say, “it worked just as well.”
A toolbox resource
“We demonstrate the importance of uncovering city-specific patterns for predicting reported crime, which provides new insight into city neighborhoods, allows us to ask new questions and evaluate policing in new ways,” says James Evans, sociologist and co-author of the study, which has been published in the journal Nature Human Behaviour.
The tool not only fine-tuned the shot geographically. When it came to “feeding” it with data, the scientists also opted for special ingredients: only homicides, assaults, injuries and robberies, crimes with which they again sought to minimize any possible deviation.
“This data was used because it is more likely to be reported to the police in urban areas where there is a historical mistrust and lack of cooperation with law enforcement. They are also less prone to enforcement bias, as is the case with drug offenses, traffic stops, and other misdemeanor offenses.
The effort to avoid bias is not accidental, nor is it a whim. The University of California algorithm is not the first to attempt to predict crime. For years there have been tools like PredPol, which uses technology to create maps with “risk areas”.
The aspiration is basically the same in all cases. With them, those responsible try to anticipate where and when it is more likely that the agents will have to face a certain crime. Other programs —COMPAS, for example— go further and jump from zoning, neighborhood maps, to the individual sphere: they assess the probability that a certain person reoffends.
However, its use has been very marked by controversy. Not only because of the thorny debate about to what extent a program can anticipate the behavior of a group, but because of biases that may interfere when making those predictions and their possible impact.
The truth is that the California study itself reveals certain deviations in the police response.
The researchers found that when crime increased in wealthy neighborhoods, police stations recorded more arrests in those areas and fewer in disadvantaged ones. If the opposite happened and it was in the poor neighborhoods where crimes increased, that rise in arrests was not reflected.
The study from @NatureHumBehav uses publicly available data on violent and property crimes, and was validated on 8 different US cities https://t.co/Wd1GzE2mam
— UChicago Biological Sciences (@ScienceLife) June 30, 2022
“When you put pressure on the system, it takes more resources to arrest more people in response to crime in a wealthy area and diverts police resources away from areas of lower socioeconomic status,” says Professor of Medicine Ishanu Chattopadhyay.
In previous tools, experts had used models in which crime emerges from “hotspots”, hotspots from which it spreads to its surroundings. For the American team, however, this approach “ignores the complex social environment of cities and does not consider the relationship between crime and the effects of police enforcement”. “Spatial models ignore the natural topology of the city,” emphasizes Evans.
The fact that the Chicago tool has been designed trying to minimize all these defects, minimizing any bias, does not mean that it is perfect. Chattopadhyay himself warns of the risks of using it to direct agents. “It should be added to a toolbox of urban policies and policing strategies to tackle crime,” he encourages.
He is not the only one who warns of the risks of recreating a particular “Minority Report” statistical version in our large metropolises. Some experts, such as Emily M. Bender, a professor at the University of Washington, advocate focusing attention and efforts on the core of the problem: address possible inequalities base and not so much to the police prediction of crimes.
Pictures | Maxim Hopman (Unsplash)
#Minority #Report #longer #science #fiction #researchers #created #algorithm #predicts #crimes