Users and Wizards in Conversations: How WoZ Interface Choices Define Human-Robot Interactions

Why Read This from an HCI Perspective

This article shows that WoZ (Wizard-of-Oz) interface design is not merely a convenient setup for experiments—it is a variable that fundamentally changes the user experience and the quality of interaction itself. In particular, it explains how the richness of nonverbal cues such as voice, gaze, and facial expressions can affect social presence, conversational connectedness, and the naturalness of the experimental interaction. This gives HCI/UX researchers important criteria for designing experiments.

CIT's Commentary

An interesting point is that the authors redefine WoZ not as a temporary workaround for ‘fake automation,’ but as a question of how well it represents future robot interactions. Limited GUIs may be advantageous for control, but they can also widen conversational gaps and amplify silence, reducing the liveliness of interaction. By contrast, VR telepresence can increase social connectedness between users and the wizard and enable denser interaction even when the operation difficulty is higher. From an HCI perspective, this suggests that interfaces are not just tools for data collection—they shape what relational behaviors ‘occur.’ Especially, the view that not only natural-language data but also contextualized nonverbal behavioral data can be treated as resources for automation offers practical implications for designing both robot learning and the experimental ecosystem together.

Questions to Consider While Reading

• Q.How well does a VR telepresence WoZ setup represent interactions with real deployed robots, and how could external validity be verified?
• Q.What interface supports would be effective in reducing the trade-off between the wizard’s high cognitive load and improved social connectedness?
• Q.When using WoZ logs that include nonverbal data for automated learning, what ethical and methodological preprocessing standards are needed?