List All Projects List Graduate Projects List Senior Design Projects List REU Projects Other Undergraduate Projects
Project # 17
ADEPT-VR: A Large Language Model - Based Adaptive De-escalation Trainer
| Team Lead: | Eshwara Prasad Sridhar |
| Other Team Members: | Jose Lopez | Anish K C | Ashirwad Sharma Dahal |
| Faculty Advisor(s): | Dr. Shuchisnigdha Deb |
| Department: | IMSE |
| Project Type: | Graduate Student |
| Title: | ADEPT-VR: A Large Language Model - Based Adaptive De-escalation Trainer |
| Abstract: | Effective de-escalation is a critical skill in modern policing. Officers frequently respond to mental health crises, suicidal individuals, and emotionally distressed subjects. These encounters are unpredictable and high-risk. Traditional training relies on role-play and scripted scenarios that are difficult to standardize, repeat, and evaluate objectively. As a result, communication skills are often practiced without measurable feedback. We developed ADEPT-VR, a fully immersive Adaptive De-escalation Trainer built in collaboration with the Fort Worth Police Department. Officers enter a virtual environment and engage in natural spoken dialogue across three operational scenarios: suicide by cop, mental health crisis, and suicidal ideation. The virtual subject is powered by a locally hosted large language model that dynamically adapts its responses to the officer’s language and approach. The system tracks conversation flow, emotional progression, response latency, and performance using structured departmental training rubrics. After each session, officers receive detailed feedback on their communication and de-escalation strategies. ADEPT-VR provides consistent, repeatable, and measurable exposure to complex crisis interactions and is being integrated into recruit training. |
| Poster: | View Poster #17 (PDF) |
| Judges Scoring: | Submit Evaluation (judges only) |