FORT WORTH, TX - The Fort Worth Police Department (FWPD) and the University of Texas at Arlington (UT at Arlington) will be collaborating on a training that is driven by artificial intelligence (AI) to enhance de-escalation techniques among its police officers.
A Department of Justice (DOJ) grant, called the Virtual Reality De-escalation Site-Based Initiative, will allow the department to supplement its traditional police training role-play with an AI-driven system to provide unpredictable, scalable and safe de-escalation training, according to the Fort Worth Report.
The more than $750,000 grant was awarded in October 2024 and the program will end in September 2027. Officer Jorge Lopez serves as the instructor for the police department, and he said this program will provide training in a safe manner, using repetitions in a controlled environment to help officers make better decisions under stressful situations.
"We want well-trained professional officers,” Lopez said. “Anything that we can drive towards that, to do it safely and efficiently and as close to realistic as possible, is always going to benefit the community because you’re going to have better trained officers."
De-escalation training for the department is typically a weeklong exercise that involves real people who are given a scripted scenario to work through. The process teaches officers how to communicate with a person in crisis and helps develop the skill set to reduce the need for use of force in critical incidents.
The AI training lasts an hour and will replace human role players with a Meta Quest 3 headset system. The system can simulate multiple situations from mental health issues or domestic violence.
The difficult of the situation can change, said Shuchi Deb, associate professor at UT Arlington's Department of Industrial, Manufacturing and Systems Engineering, while preventing liability. Data collection started this past summer with 27 volunteers from FWPD, My Health My Resources healthcare providers, and UT Arlington students who participated in recorded simulated situations.
Deb says the system utilizes the Meta headsets for immersion, the Unity game engine for scenario creation, and Gemma for conversational AI with rapid response time.
The training provides real-time feedback during the scenario and summarized constructive feedback after a module is completed for trainees to review. Virtual reality training does have its limitations. While good for procedural tasks, the system cannot teach conceptual knowledge, which will have to come from the classroom, Deb explained.
She and Lopez hope to expand the technology with multiuser scenarios for team training and introduce stress to better mimic hard-to-replicate real-world tension.
"We want it to be tested and tried and evaluated well, and then supported by the community as well,” Lopez said. The program is currently in an implementation phase as the police department is procuring all the needed logistics for their police academy.
Comments