The effects of text-based conversational teachable agents’ communicative features on student math learning: Tone style and emoji use
Bailing Lyu, Chenglu Li, Hai Li, Wanli Xing · Journal of Research on Technology in Education · 2025
As artificial intelligence (AI) continues to advance, its integration into education has generated growing interest in conversational AI as a tool to support student learning. Conversational AI systems facilitate natural language interactions between AI and students, serving diverse pedagogical roles, such as tutoring students, providing feedback, and offering learning companionship. However, most existing research has focused on conversational AI in instructor or tutor roles, which may limit students’ opportunities for independent exploration and deep engagement with learning content. Moreover, while prior work highlights the potential of conversational AI to enhance learning through personalization, engagement, and social presence, less is known about how their specific communicative features—such as tone style and emoji use—shape students’ learning experiences. To address these gaps, this study investigates conversational teachable agents (i.e., agents positioned as “students” whom learners teach) and examines how variations in their tone (positive vs. neutral) and emoji use (present vs. absent) influence students’ math learning. The findings reveal that a positive tone and the use of emojis promote students’ affective engagement during interactions with the agents, while a neutral tone and the omission of emoji facilitate deeper cognitive engagement. Furthermore, students’ cognitive engagement significantly predicted their application of procedural and conceptual knowledge during the teaching process. This study underscores the importance of communicative features in educational conversational AI and provides actionable insights for designing more engaging and cognitively supportive teachable agents.