Study: 78.8% of students use AI in shallow ways when co-programming
Only 11.1% of student-AI interactions show advanced epistemic engagement...
A new arXiv paper by Mengqian Wu introduces the concept of Epistemic AI Literacy (EAIL), reframing AI literacy as a process-oriented epistemic phenomenon that emerges during human-AI interactions. Drawing on the AIR framework (epistemic aims, ideals, and reliable processes), the study analyzes a large dataset of student-AI co-programming dialogues to detect how learners set epistemic aims and employ epistemic processes when using generative AI (GenAI) for programming tasks.
The results are striking: 78.8% of student-GenAI interactions relied on non-mastery-oriented aims and less reliable strategies such as outsourcing (letting AI do the work), verification-seeking (accepting AI outputs without deep evaluation), or prompt monitoring. Only 11.1% of interactions showed high epistemic engagement, where mastery-oriented aims (learning-focused goals) were paired with advanced strategies like epistemic justification (critically reasoning about AI-generated code). This suggests a widespread lack of deep AI literacy, with most students using GenAI as a shortcut rather than a learning tool, raising concerns about how AI is reshaping problem-solving skills in education.
- The EAIL framework detects five epistemic processes: outsourcing, explanation seeking, verification seeking, prompt monitoring, and epistemic justification.
- 78.8% of student-AI interactions used shallow strategies like outsourcing and verification-seeking, indicating low epistemic AI literacy.
- Only 11.1% of interactions achieved high epistemic engagement, combining mastery-oriented aims with advanced strategies like epistemic justification.
Why It Matters
With generative AI in classrooms, we need scalable methods to detect and foster deeper epistemic thinking in student-AI collaboration.