Exploring the Potential of a Custom GenAI-Chatbot Specialized on Diagnosis and Remediation of Alternative Conceptions on Chemical Equilibrium

Abstract Chemical equilibrium is a fundamental yet challenging concept in chemistry education and is associated with a wide range of persistent alternative conceptions. This study investigates the design and implementation of EquiBot, a customized generative artificial intelligence (GenAI) chatbot that combines diagnostic questions and conceptual change texts to support students’ structured learning processes on chemical equilibrium. During the design phase, structured testing of the chatbot performance was undertaken. For the implementation, an explorative qualitative approach was employed with six chemistry and preservice chemistry teacher students. Data collection included a chatbot interaction protocol, think-aloud session, and a semistructured interview per participant. The analysis focused on the chatbot performance and interaction process, students’ alternative conceptions, and acceptance of the conceptual change texts. The results indicate that EquiBot successfully implemented the intended interaction structure and provided targeted student support. All participants demonstrated alternative conceptions corresponding to those previously reported in the literature. Conceptual change texts were provided, and the texts were generally perceived as understandable and plausible. At the same time, persistent alternative conceptions and difficulties transferring newly acquired knowledge across contexts highlighted the complexity of conceptual change processes. Overall, the findings suggest that custom GenAI chatbots could be a promising approach to diagnose alternative conceptions in structured interaction processes and aim to remediate those conceptions by providing appropriate conceptual change texts in an automated manner.