Türkiye develops AI system to curb repeat offenses

ISTANBUL

Turkish authorities have developed an artificial intelligence-based risk analysis system aimed at assessing the risk levels of repeat offenders, specifically designed to predict whether domestic violence incidents may escalate into homicide.

This new system has its roots in an old method that came into effect in 2011, Professor Dr. Coşkun Taştan, a faculty member from the Police Academy, explained during his presentation to a parliamentary commission for violence against women.

This approach involved gathering information on the suspect as soon as a victim lodged a complaint, aiming to offer a swift procedure, but it proved insufficient in many cases.

"It was our responsibility to make this method better. We, therefore, crafted a system that enables quick, accurate and instantaneous decision-making by fortifying it with artificial intelligence,” Taştan explained.

This new application achieves this by conducting a risk analysis after key details are input into the system when a victim files a complaint. Within minutes, the AI-based system analyzes the backgrounds of both the suspect and the victim, generating a risk score.

“In other words, it warns us that ‘this could turn into murder,’ even before the crime has occurred,” Taştan noted.

Thus, if a case is classified as high risk, authorities can immediately take stricter protective measures and ensure a more swift, transparent and secure process.

The model boasts a 75 percent accuracy rate, significantly outperforming the U.K.'s HART system, another AI-based tool that assesses a suspect’s risk of reoffending (low, moderate, or high), which operates at 63.8 percent. This increased accuracy is expected to enhance law enforcement’s ability to prevent violent crimes before they occur.

Following a final test, the new system will be implemented in about 1,200 domestic violence bureaus throughout Türkiye this year, Taştan pointed out.

Beyond domestic violence cases, the system will also be utilized for general crime risk analysis.

A key issue in the justice system has been the conflict between law enforcement records and judicial decisions. Many offenders have multiple police records but remain free due to the absence of formal convictions in their judicial records. This was highlighted by the case of a 19-year-old suspect with 26 prior offenses, who was able to move freely before fatally shooting a police officer, sparking public outrage.

In an effort to address this issue, the AI model will generate risk assessment reports for courts, indicating whether an individual should be classified as a high-risk offender, helping judges make more informed decisions.