Key points are not available for this paper at this time.
This study aims to investigate how robots can express internal states through gaze control during multiparty conversations, with a particular focus on the impressions of extroversion and thinking behaviors. To this end, two robots with distinct appearances – a highly human-like android robot Nikola and a stylized, deformed-type robot CommU – were employed to assess the impact of head-to-eye ratios (HER) and gaze behavior control during thinking behaviors, on impression formation. The results of a video-based evaluation experiment firstly revealed a tendency for perceived extroversion to increase in proportion to higher HER values. Moreover, under low HER conditions, Nikola was found to elicit a stronger impression of introversion than CommU. Subsequently, an analysis based on human conversational data demonstrated that, during thinking intervals, both the duration and frequency of gaze aversion increased, whereas during recall state, upward gaze aversions were particularly pronounced. Based on these findings, a gaze control model parameterizing the thinking behaviors was developed, and its effectiveness was validated through further video-based evaluations. The results indicated that the appropriate use of gaze aversion to express cognitive processing enhances the impression of thinking behaviors, as well as the impression of extroversion. These findings suggest that the modulation of HER, combined with gaze control strategies that incorporate thinking behaviors, enables the deliberate manipulation of a robot’s perceived extroversion. The outcomes of this research are expected to support effective internal state expression across a wide range of robotic platforms, contributing to the realization of more natural and engaging human-robot interactions.
Shintani et al. (Sun,) studied this question.