SimER: AI-driven Simulation for Call-Taker Training in Emergency Responses
About SimER
SimER is an AI-driven simulation environment to assist call-taker training under emergency response scenarios. SimER is under a collaborated project supported by Vanderbilt University and Metro Nashville Government.
Fun Fact: SimER is called ANGIE at DEC to honor a beloved retired employee, Angie. There's a saying at DEC: "Ask Angie, because she knows everything."
Real-world Social Impact
What SimER Can Do
- SimER learns from past 911 calls. It is tuned using various incident specification with diverse caller images.
- SimER talks to the trainee. SimER tailors the most up-to-date LLMs as its backend agents with a sophisticatedly designed prompt generation algorithm to generate realistic, valid, and inclusive calls.
- SimER coaches the trainee, debriefing trainee performance after evaluation.
SimER Prototype Demo
Resources
Explore our research papers, code repositories, slides, and demos related to this project:
Paper Highlights
- AI-Driven Call Debriefing Revolution: First AI system automating 9-1-1 call debriefing by integrating Signal-Temporal Logic with LLMs, achieving 95.93% accuracy and saving 311.85 hours of quality assurance time.
- Solving the Coverage Crisis: Boosts quality assurance coverage from 3.32% to 85.05%, ensuring more call-takers receive timely feedback.
- Just-in-Time Feedback: Generates comprehensive reports in under 6 seconds versus 11.5 minutes for manual reviews.
- Logic-Enhanced Verification: Formalizes 2,215 call-taking requirements into STL specifications for rigorous procedural compliance.
- 92.59% User Preference: Participants cited 59.86% better actionability and 82.70% improved comprehensiveness over traditional methods.
Paper Highlights
- AI-Powered Emergency Training Revolution: First LLM-enabled training simulation for 9-1-1 dispatchers, achieving 90% user approval rating and saving 26.55+ hours of dispatcher time.
- Addressing Critical Staffing Crisis: With 25,000 unfilled dispatcher positions nationwide, Sim911 enables individual self-paced training without requiring multiple staff members.
- Real-World Accuracy & Inclusivity: Supports 57 incident types and 14 caller profiles including vulnerable populations, with 83-86% accuracy in simulating diverse demographics.
- Comprehensive Emergency Response Suite: Works with Auto311 to create an end-to-end solution for emergency and non-emergency call handling.
- Scalable Solution for 6,000+ Centers: Provides consistent, high-quality training addressing the complexity gap where traditional methods cover only 48% of incident types.
Paper Highlights
- First Automated 311 System: Pioneering automated system for non-emergency 311 calls, reducing burden on emergency services for faster 911 response times.
- 92.54% Accuracy in Call Classification: Achieves impressive F-1 score of 92.54% in predicting incident types and 93% consistency in extracting critical information.
- Confidence-Guided Intelligence: Uses advanced confidence scoring to dynamically adapt during calls and transfer complex cases to human operators when needed.
- Handles Real-World Complexity: Successfully manages multiple incident types per call (38% involve multiple issues) and leverages unintentional additional information from callers.
- Optimizes Emergency Response: Automatically generates case reports and reduces conversation turns through intelligent dialogue management while maintaining 94.49% accuracy.
Our Team
Meiyi Ma, Ph.D.
Assistant Professor in Department of Computer Science at Vanderbilt
Zirong Chen
Computer Science Ph.D Candidate at Vanderbilt
Noah Mladenovski
IT Specialist at Metro Nashville Dept. of Emergency Communications
Erin Wilson
Information Systems Application Analyst at Metro Nashville Dept. of Emergency Communications
Amber Doss
Call-taker, Dispatcher, and Emergency Response Training Specialist at Metro Nashville Dept. of Emergency Communications
Jennifer Reynolds
Call-taker, Dispatcher, and Emergency Response Training Specialist at Metro Nashville Dept. of Emergency Communications
Izeko Floyd
Quality Assurance Supervisor at Metro Nashville Dept. of Emergency Communications
Tim Watkins
Technology Manager at Metro Nashville Dept. of Emergency Communications
Kristin Mullen, RPL, CMCP
Training and Quality Assurance Manager at Metro Nashville Dept. of Emergency Communications
Paul McCallister
Deputy Director of Operations at Metro Nashville Dept. of Emergency Communications
Patrice Coleman
Deputy Director of Support Services at Metro Nashville Dept. of Emergency Communications
Stephen Martini, ENP CPE
Director of Metro Nashville Dept. of Emergency Communications
