Abstract:

This study investigates the potential of Large Language Models (LLMs), specifically ChatGPT, to elevate virtual Non-Player Character (NPC) interactions for training purposes. It focuses extensively on assessing ChatGPT's capabilities in crafting emotionally compelling conversations and responding to verbal de-escalation scenarios used for educating law enforcement personnel. The research adopts a qualitative approach, evaluating parameters like interactive storytelling, empathy, and emotional design. Additionally, it tests ChatGPT on police training protocols for verbal de-escalation. Through iterative analysis, the study reveals meaningful insights into creating efficient LLM configurations to simulate human-like exchanges for targeted training. The findings showcase the successes and shortcomings of using ChatGPT for basic law enforcement education, particularly in comprehending fundamental de-escalation techniques. The research highlights promising opportunities for developing AI-based education-training tools focused on communication-heavy fields. It advocates continued LLM exploration to elevate training conversations and NPC interactions across high-impact domains.

Keywords:

Large Language Models, Virtual Agents, Law Enforcement Training, Non-Player Characters.