A Chatbot Won't Judge Me: An Exploratory Study of Self-disclosing Chatbots in Introductory Computer Science Classes

Students in introductory Computer Science (CS) courses sometimes struggle with learning course content, but feel these struggles are uniquely theirs. To foster a more inclusive CS culture and normalize challenges in the learning process, we designed a conversational agent ("chatbot") that self-discloses information about the chatbot's own imaginary struggles with learning course material. Inspired by previous work in the mental health domain where humans reciprocated disclosure when a chatbot disclosed sensitive information, our goal was to promote student self-disclosure of learning challenges and to help students feel less alone. To inform design, we first conducted three focus groups with CS students on themes of identity and belonging. Based on these findings, we designed a self-disclosing chatbot ("Mibi") and deployed it in a pilot summer course (40 students) and a larger course (460 students) in the fall semester of 2023. Our work is the first real-world deployment of a chatbot in higher education for promoting student wellbeing, rather than assisting with practical course content. We highlight findings from this exploratory study, sharing how students engaged with Mibi, where it succeeded, where it has room to grow, and how that can inform future iterations of this promising new classroom companion for student mental health.