Abstract
Virtual humans (VHs) are human-like intelligent agents increasingly used in human–artificial intelligence (AI) collaborative contexts, where control lability is essential for fostering user trust and ensuring a human-centered experience. Although widely adopted, VHs products are often reported to suffer from poor control lability due to the gap between the limited capabilities of narrow AI and users’ high expectations shaped by their anthropomorphic appearance. However, existing control lability theories focus primarily on action-level mechanisms such as undoing and redoing, providing limited support for users’ perceptual and cognitive understanding of interacting with and controlling AI-infused products. To bridge the expectation-reality gap, this study revisits Bill Verplank’s interaction design questions and reframes control lability into a triadic framework centered on the questions: (1) How do users feel in control? (2) How do users know the pathway to control? (3) How do users do the control actions? Using grounded theory, we derived the triadic framework from existing design guidelines and recent research. We proposed a systematic design approach to enhancing control lability in VHs, comprising nine design principles and three interaction modules corresponding to the feel, know, and do phases. The approach was implemented in an LLM-based VH system at a technology experience center to evaluate practical applicability. An empirical study compared two systems, with and without the design approach. Results show that the approach facilitates control lability beyond mere execution of control actions, enabling users to understand AI better during human–virtual human interaction (HVHI), thereby promoting a more human-centered AI experience.
Keywords
Controllability; Virtual human; Interaction design; Human-centered AI
DOI
https://doi.org/10.21606/iasdr.2025.507
Citation
Xiao, L., Wu, Q., Xie, Y., Chen, M., Zhang, Z., Guo, X., Zhu, Y., Li, Y.,and Dong, Y.(2025) Bridging the Expectation–Reality Gap: A Systematic Design Approach to Facilitate Controllability in Human–Virtual Human Interaction, in Chang, C.-Y., and Hsu, Y. (eds.), IASDR 2025: Design Next, 02-05 December, Taiwan. https://doi.org/10.21606/iasdr.2025.507
Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License
Conference Track
Track 4 - Human-Centered AI
Bridging the Expectation–Reality Gap: A Systematic Design Approach to Facilitate Controllability in Human–Virtual Human Interaction
Virtual humans (VHs) are human-like intelligent agents increasingly used in human–artificial intelligence (AI) collaborative contexts, where control lability is essential for fostering user trust and ensuring a human-centered experience. Although widely adopted, VHs products are often reported to suffer from poor control lability due to the gap between the limited capabilities of narrow AI and users’ high expectations shaped by their anthropomorphic appearance. However, existing control lability theories focus primarily on action-level mechanisms such as undoing and redoing, providing limited support for users’ perceptual and cognitive understanding of interacting with and controlling AI-infused products. To bridge the expectation-reality gap, this study revisits Bill Verplank’s interaction design questions and reframes control lability into a triadic framework centered on the questions: (1) How do users feel in control? (2) How do users know the pathway to control? (3) How do users do the control actions? Using grounded theory, we derived the triadic framework from existing design guidelines and recent research. We proposed a systematic design approach to enhancing control lability in VHs, comprising nine design principles and three interaction modules corresponding to the feel, know, and do phases. The approach was implemented in an LLM-based VH system at a technology experience center to evaluate practical applicability. An empirical study compared two systems, with and without the design approach. Results show that the approach facilitates control lability beyond mere execution of control actions, enabling users to understand AI better during human–virtual human interaction (HVHI), thereby promoting a more human-centered AI experience.