This post is also available in: עברית (Hebrew)
Artificial intelligence (AI) technologies can now be used to expose the truth of a fake claim filed for insurance purposes without wasting police time.
Law enforcement agencies across Spain have adopted an artificial intelligence (AI) system capable of uncovering fake crime and theft claims.
The AI system, known as VeriPol, was developed by researchers from Cardiff University and the Charles III University of Madrid. The system uses automatic text analysis and machine learning to identify false statements.
According to the computer scientists, VeriPol is able to identify false robbery reports “with over 80 percent accuracy.”
When it comes to issues such as filing false robbery statements to the police, a claim can then be issued against insurance providers to fraudulently benefit from a policy.
Law enforcement budgets and staff are often strained and so it can be difficult to prove a suspicious statement is false, as well as find the time and resources to do so.
The AI tool is able to analyze written statements to recognize patterns which are most commonly associated with false statements, such as the kinds of items allegedly stolen, the descriptions given of supposed attackers, and the “finer points” of an incident.
VeriPol uses what is called natural language processing, an element of machine learning technologies which helps artificial systems understand and interpret the natural use of human language.
Algorithms are used to decode this language, and VeriPol’s understanding is based on historical police reports which have been fed into the system.
In addition to helping police staff decide which claims to allocate resources towards, VeriPol may deter members of the general public from filing false statements in the first place, saving agencies both time and money.
VeriPol has been rolled out across law enforcement agencies in Spain. A trial launched by the Spanish National Police used VeriPol to scrutinize over 1000 police reports, in which the tool could detect a false report eight out of 10 times, according to zdnet.com.