Fri. Apr 19th, 2019

The artificial intelligence of the Police that exposes false accusations

La inteligencia artificial de la Policía que desenmascara denuncias falsas


It started as a simple experiment at the end of 2015 and in just over three years it has become the most effective tool based on the artificial intelligence of all those who handle the police around the world. Developed by an inspector of National Police Force on the way back from a doctorate course in Predictive Police in the University of California Los Angeles, VeriPol has reached a 96% of correctness in the detection of false accusations in patrimonial crimes. No other tool achieves this level of efficiency.

Sponsored Ads

Advertise Here

The majority of these complaints are for Stole or theft of mobile phones without intimidation, breakage of moons in cars or high sums of money that, coincidentally, the complainant was carrying. And the whistleblowers do not pursue anything other than cheating an insurance company or trying to explain an unspeakable expense. Although sometimes there are also cases of incomprehensible assaults to domiciles in which all kinds of goods disappear, generally also insured.






Developed by a body inspector, this tool is already 96% accurate

Although beyond hunting liarVeriPol has become the best deterrent mechanism. Since he arrived at all the police stations of the National Police last October-he is not yet in Catalunya neither Euskadi, where the police themselves have competences in citizen security - this type of fraudulent complaints has dropped by 40%. Although the police do not want to attribute all the merit to the tool, many complainants are repetitive or their environment is and now think twice has much to do with its implementation.

VeriPol works with a algorithm who picks up some 300 variables and is based on the natural language. Basically, because when denouncing a real and a figurative robbery, the complainants use specific and coinciding formulas in both cases. "Everyone tells the truth with a similar argument and lies in the same way, in the same way that in a case they offer some details that on the contrary tend to ignore - explains Miguel Camacho, one of the drivers and developers of the tool. The tool detects it, establishes a percentage of truthfulness and the agent that collects the complaint soon falls in contradiction to the false reporter alerted by VeriPol. "






Everyone tells the truth with a similar argument and lies in the same way; the algorithm detects it

To the formulation of the complaint, logically, other elements are added that the algorithm also ponders to formulate its resolution. And with the data that is collected and the resolution of the complaints VeriPol also learn through what could be called a mechanism of
machine leaning
as elementary as effective. "When we did the first pilot tests in Malaga and Cartagena we confirmed that 90% success of which we started with the analysis of cases already closed, now we go for a 96%", explains the commissioner Francisco Álvarez, operational manager of the tool.

Miguel Camacho, developer of VeriPol

Miguel Camacho, developer of VeriPol
(RTVE)

The secret of VeriPol, according to its developer, is none other than that simplicity of its formulation. Common sense led to an algorithm and the accumulation of data that the most experienced police can not process. The tool began with an exhaustive emptying of all the elements that Camacho and a team of experts in mathematics and
big data
of the Complutense University, the Carlos III University Y The Sapienza of Rome considered ponderables of 534 true and 588 false allegations. To this he added the same comparable characteristics in the opinions of Tripadvisor identified as true and false. And VeriPol surprised both locals and strangers in their first analysis.





"The efficiency of VeriPol, a tool developed exclusively in Spain, surprised all specialists in artificial intelligence and we could say that it has marked a before and after in the development of criminological tools", explains Fernando Miró, director of Crímina Research Center for the Study and Prevention of Crime, pioneer in the development of artificial intelligence tools aimed at criminal, police and judicial investigation.

Although VeriPol is not, far from it, the panacea. Fundamentally because, as its creators recognize, it has the advantage that its object of analysis is often coarse and often improvised deception. False whistleblowers do not usually seek anything other than to cheat Insurance company or justify a high expense. And even though they may be repetitive in their artifice, they are not professional criminals in any case. In short, it is easy to catch them. Especially now.





On the positive side of the balance is not only the development of a tool that continues to be a spearhead, a pioneering project, but its already proven deterrent and, above all, its ability to simplify police investigations and dedication of resources to a type of denunciations that today are still the most common. "The police are obliged to investigate everything and VeriPol will never decide, but it takes away a lot of work, that is undeniable," Álvarez concludes.







Source link

Leave a Reply