A systematic review methodology was used to review publications on the emerging use of AI chatbots in science classroom, using the traditional and refined model of science identity. The traditional model of science identity entails how an individual recognises himself or herself as a science person, while the refined model of science identity more ambitiously considers how an individual was being recognised as doing science by AI chatbots and their peers. Thus, the traditional model of science identity entails interest, performance/competence and recognition. Whereas in the refined model, science identity is defined as entailing disciplinary epistemic affect, students’ self-competence in scientific practices as well as recognition as a “science person” by peers and chatbots. This review aims to identify strengths and weaknesses of 19 studies in terms of training AI chatbots, technical pedagogical design, targeted outcomes and changes in these outcomes. Results show that AI chatbots in these studies mostly target outcomes and pedagogical design regarding students’ interest in science, while their attention to disciplinary epistemic affect, students’ self-competence in scientific practices and recognition as a “science person” is minimal. Hence, this review proposes incorporating the refined model of science identity into pedagogical design, teacher education and curriculum development in the use of AI chatbots in science education.
Kason Ka Ching Cheung (2026) studied this question.