Amrita Ganguly, Sai Sharanya Garika, Aditya Johri (2026) — arXiv preprint.
This study investigates how students conceptualize generative AI (GenAI) and how those mental models shape their academic integration. A student's mental model of GenAI — their beliefs about what it can and cannot do — influences both perceived capability and choices about when to delegate tasks. The authors surface the range of student conceptions, from tool-as-calculator to collaborator, and show that inaccurate or shallow models correlate with over-reliance and weaker learning outcomes.
The work connects to broader debates on ai-literacy and student-experience with AI, arguing that mental-model accuracy is a prerequisite for productive human-ai-collaboration. It extends metacognition research by treating AI understanding as a learnable metacognitive skill, and bears on generative-ai use in higher-ed. Implications include designing interventions that explicitly calibrate students' models rather than assuming fluency.
Related Pages
- ai-literacy — AI literacy as a learnable competency
- student-experience — How students experience and integrate GenAI
- metacognition — Mental models as metacognitive skill
- over-reliance — Shallow models drive over-reliance
- human-ai-collaboration — Prerequisite for productive collaboration
- generative-ai — Core technology under study