Role priming of AI conversational agents: Effects on language learners’ emotions and engagement

Abstract

This study examines how simple role framing, introducing a general-purpose AI as a language tutor, conversation buddy, or language tool, affects learners’ emotional responses, perceived interaction styles, and engagement in AI-mediated language learning. Drawing on the Computers Are Social Actors (CASA) framework, we conducted a between-subjects experiment with 213 Chinese university-level EFL learners. Participants engaged in short AI conversations under one of the three framing conditions, followed by a structured questionnaire assessing emotions, engagement, and perceived AI roles. Learners in the buddy condition reported significantly greater emotional comfort, including reduced anxiety and increased confidence, compared to the tutor and tool groups. Role framing also shaped learners’ perception of the AI’s interpersonal stance and influenced their self-reported engagement. Correlation and regression analyses confirmed that emotional experience predicted engagement, while mediation analysis revealed a small but significant indirect effect of AI role on engagement via emotion. These findings suggest that role framing can shape learners’ affective and behavioral responses even when the AI’s linguistic output remains unchanged. This highlights the importance of socio-emotional design in AI-assisted learning and suggests that minimal psychological cues can meaningfully enhance learner experience.