For decades, Elia Lombardo hunts criminals through the streets of Naples and other parts of Italy. And it has been combining the study of behaviors, characteristics, customs and strategies of scammers, pickpockets, thieves and serial robbers with that of computer programming languages for more than 20 years. Thus, this superior inspector of the Italian Police has been able to develop XLAW, an algorithm capable of predict crimes Street with precise indications about place, time, geographical area and type of author and victims. The police have already used this system for a long time in Naples and Prato. This month he also implemented it in Venice, where, a few days after its start-up, he allowed to stop infraganti to a man who robbed in a hotel.
Lombardo, 51, says the key to developing the algorithm was a paradigm shift when trying to understand the mechanisms of crime. "Before I used statistical systems. It located the areas most affected by a certain type of crime and sent patrols there. It is the principle that is still used in almost everyone, "he explains. But the results in crime prevention were not what I was looking for. So this agent graduated in Communication Sciences enrolled in university courses in criminology, and things went better.
"I understood that some crimes, such as robberies, robberies and robberies, are predictable because they are routine and usually committed by professionals," says Lombardo. The inspector details that criminals of this type use the same strategies at all sites. "They choose areas where there are more" dams "and escape routes or adequate shelters. I define hunting reserves. "
In addition, these delinquents "usually plan their criminal activity according to a monthly temporary scheme, like any worker," the police adds. To discover his methods better, he also looked at what moments they choose to act. "In most cases they are cyclical phases of operation, such as the arrival of a train or the opening and closing hours of stores," he explains. According to it indicates, more than 85% of the crimes that are committed in a certain territory, in spite of which it is, have these characteristics.
An algorithm based on criminal models
The study of these dynamics, which Lombardo defines "criminal models", is the basis of the XLAW algorithm. In it, data corresponding to complaints and police interventions are inserted. "Between the crimes that are verified in a territory, the program indicates those that belong to these models" he assures. The system launches predictive alerts, which are displayed with different colored circles on a map thanks to a GIS technology [Sistema de Información Geográfica]. "They give information such as the schedule and the street in which there will be a robbery, the age range and the form of action of the alleged perpetrator and the type of victim," he explains.
Thus, those responsible for the operational units obtain clues to prevent criminal acts planned in the next two hours and can give specific instructions to the patrols on how to distribute in a territory. The same agents on the street can view alerts on a mobile device.
Lombardo began to develop the algorithm in 2004. Seven years later, the police extended its use to the entire territory of Naples. "In these years, crimes of this type have been reduced by 24%. And the denunciations and detentions infraganti have increased by 27%, "he says. In 2017 the system was also implemented in Prato, a city in Tuscany where the percentage of crimes last year had increased, according to a trend contrary to the national average.
There, the Neapolitan inspector maintains, the results of his creation were even better. Since it was used, street crime has been reduced by 39% and complaints and arrests have increased by 54%, according to the data provided by Lombardo. The success of XLAW seems unstoppable. At the beginning of November its implementation in Venice was announced. At dawn on the 16th, a call warned that a thief had entered to rob in a hotel in Mestre, the land area of the city. A patrol alerted in advance thanks to the algorithm was already waiting for him at the door before he left.
Increasingly, technology conquers space as a companion to police forces. Different companies market systems destined to increase the possibility of predicting crimes. Among the most popular tools is PredPol, a software that generates geographic alerts of possible new crimes according to the probability of risk. Another is COMPAS, an algorithm that calculates the probability that a person will repeat the same crime that he or she committed and more than once considered controversial. because of the possibility that it is based on racist biases or is not more effective than humans.
In Spain, the National Police and the Civil Guard do not use systems such as XLAW, according to sources from the Ministry of the Interior. "The study of the integration of algorithms of this type in these bodies is still at a level of academic work," the same sources add. On the other hand, the National Police launched this yearA pioneering artificial intelligence system to detect false allegations. Sources of this body assure that a computer system is also used that is based on statistical data to indicate which crimes are more common in each zone.
Inspector Lombardo says that his algorithm, when detecting future crimes, does not use any harmful bias from the people involved. "The machine does not focus on the subjects and their biometric characteristics, but on the way they operate". The software can autonomously incorporate new criminal models. "When I developed it, in Italy there was no phenomenon of baby gangs [bandas criminales callejeras formadas por individuos muy jóvenes, en algunos casos menores]. Towards the year 2016, the algorithm took that model alone, "he says.
In the opinion of this policeman, XLAW can also be useful for other types of crimes, such as acts of vandalism or abusive waste dumping. "We are doing tests, and we also want to work on the issue of gender violence and harassment," he adds. The efficiency of the system is convincing the Italian Ministry of the Interior to extend its use throughout the national territory, says Lombardo.
The Eurocop Chair, a project of the University Jaime I (UJI) of Castellón and the company Eurocop, has developed a program of prediction of crimes based on a probabilistic calculation model, different from that of XLAW. The system is being tested by the local police in Castellón.
"Our software it produces maps of risk forecast of a crime in certain places of a city and in certain moments of the day ", explains Jorge Mateu, director of the chair and teacher of the UJI.