User-Chatbot Similarity in Communication Style and Avatar Appearance for Enhanced Learning in Education

This study examined how user-chatbot similarity in communication style and avatar appearance affects university students’ learning outcomes, and whether trust in the chatbot mediates this effect. Using a 2 × 2 between-subjects design (N = 88), participants interacted with a chatbot similar or not similar in communication style and appearance. Communication style similarity was achieved by analyzing participants’ written input to generate a linguistic profile, which guided chatbot responses using ChatGPT. Appearance similarity was manipulated using a detailed avatar creator that let participants design self-resembling avatars. Contrary to expectations, the study did not reveal significant effects. Yet, it contributes to a deeper understanding of the dynamics of perceived human-chatbot similarity in digital entities. A post-study validation (N = 21) and post-experiment interviews (N = 29) uncovered potential barriers to user self-perception in relation to digital agents. These findings highlight the complexity of facilitating self-identification with digital agents and point to valuable directions for future research.