Embodied Copresence Confers No Advantage Over Screen-Based Delivery in Robot-Facilitated Language Learning

The past decade has seen growing interest in deploying social robots as assistive technologies in educational settings. Key unresolved questions concern the extent to which, and possibly when, the embodied copresence of a robot confers benefits over screen-based alternatives. This is an issue with applied significance given the logistical and cost constraints of deploying physical robots at scale. In a preregistered between-subjects experiment, 40 adults completed a Spanish language learning task alongside a Pepper humanoid robot that was either physically co-located next to participants (Embodied, n = 20) or presented via live video feed on a large screen next to participants (Screen, n = 20). Contrary to our predictions, conditions did not differ on any objective measure of task engagement (accuracy, completion time, voluntary continuation), self-report indices (enjoyment, difficulty), attention allocation to the robot or task (subjectively coded gaze orientation), or validated measures of robot perception and affiliation (Godspeed questionnaire; Self-in-Other scale), with the exception of greater perceived task difficulty and robot awareness in the Embodied than Screen condition. Reflexive thematic analysis of post-task interviews identified three cross-condition themes: robot learning companions (1) create experiences of invigilation; (2) are expected to fulfil multiple social roles (motivational coach, tutor, and companion); and (3) are expected to interact in a human-like, intelligent, and adaptive manner. These findings suggest that, for adult learners engaged in short-duration tasks, such as the simple language-learning task explored here, screen-based delivery may be functionally equivalent to physical co-presence, with implications for the scalable deployment of social robots in educational contexts.