Omar Shaikh

Comparing Text-Only and Virtual Reality-Embodied Conversational AI Agents for Interpersonal Skills Training

Yejoon Yoo
Andrea Stevenson Won
Companion Publication of the 2025 Conference on Computer-Supported Cooperative Work and Social Computing (CSCW), 2025

Abstract

Conversational AI agents powered by large language models (LLMs) have the potential to support the development of interpersonal skills, which are essential for navigating diverse situations and engaging effectively with a variety of people. However, text-based AI agents often lack crucial nonverbal cues such as facial expressions, body gestures, and tone of voice. In this study, we present a VR simulation featuring an embodied AI agent that leverages nonverbal cues to train interpersonal skills across various scenarios. We compare its efficacy to a Text-Only AI agent in a between-subjects study with twenty-four participants. We find that participants preferred the embodied agent condition, and their initial scores were significantly higher than those of participants in the text condition. However, the difference between the initial and final scores was not statistically significant.

Materials

Project
PDF

BibTeX

			
@inproceedings{10.1145/3715070.3749224,
  author = {Yoo, Yejoon and Sun, Yilu and Shaikh, Omar and Won, Andrea Stevenson},
  title = {Comparing Text-Only and Virtual Reality-Embodied Conversational AI Agents for Interpersonal Skills Training},
  year = {2025},
  isbn = {9798400714801},
  publisher = {Association for Computing Machinery},
  address = {New York, NY, USA},
  url = {https://doi.org/10.1145/3715070.3749224},
  doi = {10.1145/3715070.3749224},
  abstract = {Conversational AI agents powered by large language models (LLMs) have the potential to support the development of interpersonal skills, which are essential for navigating diverse situations and engaging effectively with a variety of people. However, text-based AI agents often lack crucial nonverbal cues such as facial expressions, body gestures, and tone of voice. In this study, we present a VR simulation featuring an embodied AI agent that leverages nonverbal cues to train interpersonal skills across various scenarios. We compare its efficacy to a Text-Only AI agent in a between-subjects study with twenty-four participants. We find that participants preferred the embodied agent condition, and their initial scores were significantly higher than those of participants in the text condition. However, the difference between the initial and final scores was not statistically significant.},
  booktitle = {Companion Publication of the 2025 Conference on Computer-Supported Cooperative Work and Social Computing},
  pages = {194–198},
  numpages = {5},
  keywords = {Embodied AI Agent, Large Language Models, Virtual Reality, Interpersonal Skills Training, Conflict Resolution, Human-AI Interaction, Nonverbal Communication},
  location = {},
  series = {CSCW Companion '25}
}