TLDR: New research from North Carolina State University indicates that integrating ’embodied’ AI agents within virtual reality (VR) environments significantly enhances programming students’ confidence and engagement. Unlike voice-only AI, AI peers with a virtual body that can interact with the environment are more readily accepted as partners, fostering a more effective learning experience.
Raleigh, NC – A groundbreaking study conducted by researchers at North Carolina State University has revealed that ’embodied’ artificial intelligence (AI) agents in virtual reality (VR) settings can dramatically improve programming students’ confidence and engagement. The findings suggest that giving AI peers a physical presence within the virtual environment, allowing them to interact and gesture, makes them more effective learning partners compared to their voice-only counterparts.
Published on October 1, 2025, the proof-of-concept study explored the impact of different AI agent types on student learning in VR. Qiao Jin, an assistant professor of computer science at NC State and the corresponding author of the paper, highlighted the novelty of the approach. “Using AI agents in a VR setting for teaching students programming is a relatively recent development, and this proof-of-concept study was meant to see what kinds of AI agents can help students learn better and work more effectively,” Jin stated.
The research focused on ‘pAIr’ learning, a concept where an AI agent acts as a programming peer. Jin emphasized the significance of the results, noting, “Peer learning is widespread in the programming field, as it helps students engage in the learning process. For this work, we focused on ‘pAIr’ learning, where the programming peer is actually an AI agent. And the results suggest that embodying AI in the VR environment makes a real difference for pAIr learning.”
The study involved 18 adult participants interested in programming, 14 of whom had minimal or no prior experience. Participants engaged in two programming tasks with support from an embodied AI agent – one structured and one unstructured. They also completed a separate structured task with a voice-only AI agent.
Post-task interviews provided crucial insights. While both AI versions supported learning, the embodied agents stood out. Students reported feeling more engaged with the pair learning process when interacting with the embodied AI, which in turn boosted their confidence and motivation. Furthermore, the gestures made by the embodied AI, such as pointing to specific sections of code on the screen, significantly improved students’ ability to understand instructions.
Also Read:
- AI Competency Gap: Education Sector Faces Skills Shortfall Amidst Rapid Technological Integration
- Cornell Faculty Navigate Dual Path: Integrating and Restricting AI in Education
The researchers concluded that embodiment improved engagement, enhanced communication efficiency, and offered emotional support. The incorporation of embodied actions allowed users to perceive the AI agent as a genuine ‘programming partner,’ facilitating interactions akin to those shared with real-life collaborators. The study underscores the potential of embodied AI agents to advance programming education and offers valuable design insights for creating more intuitive and supportive AI programming partners, though it also noted that the effectiveness of embodied actions is dependent on their timing and accuracy.
