Fort Worth police, UT Arlington use AI, VR for de-escalation training

The Fort Worth Police Department and the University of Texas at Arlington are collaborating to integrate artificial intelligence and virtual reality into officer de-escalation training. The initiative, funded by a Department of Justice grant, aims to supplement traditional role-play exercises with a scalable and safe digital system.

The Virtual Reality De-escalation Site-Based Initiative received more than $750,000 in funding in October 2024. The program is scheduled to conclude in September 2027. Traditional de-escalation training typically involves a weeklong exercise where officers interact with human role players in scripted scenarios to practice communication and reduce the need for force.

In contrast, the new AI-driven module lasts approximately one hour and replaces human actors with a Meta Quest 3 headset system.

Officer Jorge Lopez, who serves as the instructor for the department, stated that the program allows for repetitive practice in a controlled environment. This approach helps officers make better decisions under stress. Lopez emphasized that the goal is to produce well-trained professionals, noting that safe and realistic training benefits the community by improving officer preparedness.

Shuchi Deb, an associate professor at UT Arlington’s Department of Industrial, Manufacturing and Systems Engineering, explained that the system can simulate various situations, including mental health crises and domestic violence incidents. The difficulty of these scenarios can be adjusted while preventing liability issues. The technology utilizes the Unity game engine for scenario creation and Gemma for conversational AI with rapid response times.

Software development took a year to properly convert audio to text and enable the computer to respond via text-to-speech.

Data collection began this summer with 27 volunteers from the police department, My Health My Resources healthcare providers, and UT Arlington students. These participants engaged in recorded simulated situations. The system provides real-time feedback during the scenario and a summarized constructive review after each module is completed.

Deb acknowledged that while virtual reality is effective for procedural tasks, it cannot teach conceptual knowledge, which must still be learned in a classroom setting. She and Lopez hope to expand the technology to include multiuser scenarios for team training and introduce stress elements to better mimic real-world tension.

Lopez advised that law enforcement should carefully evaluate new technologies before investing in them, ensuring they are tested and supported by the community. The program is currently in an implementation phase as the department procures logistics for the police academy.

Source: fortworthreport.org.

Sources

fortworthreport.org