Analysis of Mutual and Referential Human and Robot Gazes in a Collaborative Word Association Game

This study investigates how a robot's gaze behavior influences human visual attention during a cognitively demanding, collaborative word association game. While previous research has extensively studied gaze in social, face-to-face conversations, little is known about how gaze functions in task-focused interactions where attention is split between a partner and a shared environment. The researchers used a NAO robot as an LLM-driven partner and compared two conditions: one where the robot primarily used mutual gaze (looking at the participant) and one where it used referential gaze (looking at the game board).

The researchers analyzed gaze data using eye-tracking glasses and Areas of Interest (AOI) segmentation. They found that the robot's gaze orientation had no significant impact on the time it took for participants to first fixate on a word proposed by the robot. Furthermore, while participants tended to look at the robot more often when making confirmation requests, this behavior was not significantly influenced by the robot's gaze condition. The authors conclude that in high-demand tasks, the verbal and task-related requirements of the game likely overshadow the communicative effects of the robot's non-verbal gaze cues.

Understanding how humans and robots coordinate attention is critical for designing effective collaborative systems, such as those used in assembly or assistive healthcare. This study suggests that in complex, task-oriented environments, robots may not need to prioritize sophisticated gaze-cueing behaviors to maintain task efficiency, as human attention is naturally anchored to the task itself. These findings provide a baseline for future research into how cognitive load and task structure modulate the effectiveness of social cues in human-robot interaction.