Developing and Evaluating the Fidelity of Virtual Reality-Artificial Intelligence (VR-AI) Environment for Situated Learning

April TanMichael C. Dorneich^*Elena Cotos

• Iowa State University, Ames, United States

This study develops and evaluates a Virtual Reality-Artificial Intelligence (VR-AI) learning environment designed to facilitate the socio-rhetorical socialization of graduate students. Academic discourse socialization is crucial, yet traditional classroom instruction often lacks authentic socialization opportunities, limiting students' exposure to their disciplinary communities. To address this gap, this study integrates genre and situated learning theories to create an immersive VR-AI environment that simulates academic conference poster sessions. Learners interact with AI-driven agents acting as expert researchers and mentors, engaging in discussions and receiving real-time feedback on research communication. The study focuses on developing, operationalizing, and evaluating the fidelity of the VR-AI environment across four key dimensions: physical, functional, psychological, and social fidelity. Twenty participants tested the environment, completing two learning tasks: engaging with poster presenters and reflecting with a major professor. Fidelity was assessed using mixed methods, including presence questionnaires, workload assessments, behavioral observations, and semi-structured interviews. Results indicate high physical and functional fidelity, with participants describing the environment as immersive and reflective of real-world academic settings. Psychological fidelity was also well represented, as learners engaged in cognitively demanding research discussions and rhetorical reflection. However, social fidelity remained a challenge, as AI agents struggled with conversational turn-taking and response length, reducing the authenticity of academic exchanges. These findings highlight the potential of VR-AI environments for disciplinary socialization while underscoring the need for refined AI-driven interaction designs to support more fluid, reciprocal dialogue.