Secondary Students' Emerging Conceptions of AI: Understanding AI Applications, Models, Engines and Implications

Despite the growing use of AI technologies, there is limited empirical evidence of secondary students’ understanding of AI. Gaining insights into ways in which this might be influenced by their everyday engagement with these tools can support the teaching of AI, including the design of educational resources. The SEAME framework proposes four dimensions for thinking about AI: (i) the social and ethical (SE) implications of AI; (ii) its use in applications (A); (iii) how models (M) are trained; and (iv) the underlying engines (E) used. In this paper, we investigate students’ emerging conceptions of AI and how these can be understood through classification using an appropriate framework. Drawing on data from 474 secondary and college students (11–18), we found that students mostly focused on the applications of AI, followed by its social and ethical implications. We found evidence that students held some primarily accurate conceptions of how AI works, such as how models are trained using large datasets and how human behaviours are simulated (e.g. talking, problem-solving). Specific conceptions relating to generative AI tools were also observed, including the ability to generate content and response to prompts. Conversely, we found students held naive conceptions relating to AI systems having agency or emotions. We argue that efforts to promote accurate conceptions of AI should build on students’ conceptions and take into account how language and representation are used (e.g. anthropomorphism). The results of this study have implications for educators, resource developers and researchers.