Can LLMs Generate Behaviors for Embodied Virtual Agents Based on Personality Traits?
Bin Han, Deuksin Kwon, Spencer Lin, Kaleen Shrestha, Jonathan S. Gratch · 2025
This study proposes a framework that uses personality prompting with Large Language Models (LLMs) to generate verbal and non-verbal behaviors for virtual agents based on personality traits.Focusing on extraversion, we evaluated the system across two scenarios-negotiation and ice-breaking-using both introverted and extroverted agents.In Experiment 1, we ran agent-agent simulations and conducted linguistic analysis and personality classification to assess whether the LLM-generated language reflected the intended traits, and whether the corresponding nonverbal behaviors differed by personality.In Experiment 2, we conducted a user study to evaluate whether these personality-aligned behaviors were consistent with their intended traits and perceptible to human observers.Our results show that LLMs can generate verbal and nonverbal behaviors that align with personality traits, and that users are able to recognize these traits through the agents' behaviors.This work highlights the potential of LLMs in shaping personality-aligned virtual agents.