Learning from Novice Programmers’ Metaphors of Generative Artificial Intelligence

Abstract

Generative artificial intelligence (GenAI) tools are reshaping how students approach problem-solving in programming education. This study demonstrated the methodological value of metaphor analysis in examining students’ perceptions of AI-supported learning. It explored how first-year students in an introductory Python programming module at a South African university conceptualize GenAI use during coding-related tasks. The context matters because many students from disadvantaged communities enter university with limited computing exposure. Guided by a framework of AI-mediated cognitive engagement, the study drew on 161 responses and used inductive-deductive metaphor analysis to examine how students position AI in relation to their thinking. Six dominant metaphorical framings emerged: tool, tutor, facilitator, specialist, liar, and threat. These metaphors revealed that students experienced AI in both constructive and cautionary ways, as a source of cognitive support and conceptual expansion, and as a potential source of reduced mental effort. The findings showed a systematic correspondence between metaphorical framings and reported cognitive experiences, including mind-mending, mind-stretching, mind-gaming, and mind-replacement. Chi-square and Fisher’s exact tests strengthened the qualitative findings by confirming an association between metaphor roles and cognitive influence. The study contributes to research on AI in education by showing that metaphor offers a powerful lens for understanding how students interpret and regulate AI-supported learning. It argues that GenAI is a conditional cognitive mediator whose impact depends on student agency, epistemic trust, and the nature of engagement. The findings have implications for the design of programming curricula and educational practices, integrating AI responsibly while sustaining critical thinking and academic integrity.