How Embodiment Shapes Human–AI Interaction: Evidence from Real-World Deployment of Holographic and Screen-Based Systems

Abstract

Artificial intelligence (AI) is increasingly moving into physical environments through kiosks, avatars, and holographic systems. While significant research has focused on model performance, less is understood about how the form of AI systems influences human interaction. This study examines the impact of embodiment by comparing user interactions with a holographic AI assistant and a screen-based avatar deployed at CES. Observational and interaction data reveal that the holographic system consistently elicited longer engagement, more exploratory questioning, and greater user interaction than the screen-based system. Notably, these differences occurred despite both systems using the same underlying AI model. The findings suggest that embodiment shapes user behavior independent of system functionality. This work contributes a practical framework for evaluating embodied AI systems and highlights the importance of considering physical form in the design of future human–AI interactions. This work presents a behavioral evaluation framework for embodied AI systems that leverages interaction time, question style, user behavior, and language use.