AI-section students showed tool-linked self-efficacy and attenuated metacognitive calibration, with the human-pair contrast group showing larger concept-map gains in cross-links and hierarchy¶

Claim · All claims
Evidence · 1 study (2 entries) · qualitative r3 · q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q2 i? Fourteen of 24 AI-user students described growing pride in prompt craft, sometimes conflating Tool Mastery with programming competence. → Liu 2026 q2 i? Indicators of attenuated metacognitive calibration—mismatch between perceived readiness and independent capability—co-occurred with frequent delegation, and the human-pair contrast group showed larger descriptive increases in concept-map cross-links and hierarchical organization. → Liu 2026 (2)

Evidence¶

Liu 2026¶

Liu, D., Fan, G., & Pan, L. (2026). Tool, tutor, or crutch?: A grounded theory of cognitive scaffolding and offloading in AI-assisted programming education. International Journal of STEM Education. https://doi.org/10.1186/s40594-025-00592-w

q2 · i? · qualitative · r3

Dyadic interview analysis of the AI-user group: 14 of 24 students described pride in writing perfect prompts, with self-efficacy tethered to managing the tool rather than reasoning from first principles. A contrasting case (Peng) used prompts as scaffolds for reasoning with test-first edits.

"A recurring theme in the interviews with more than half of the AI-user group (14 of 24) was a developing pride in their “prompt craft.”"

Liu 2026 (2)¶

Liu, D., Fan, G., & Pan, L. (2026). Tool, tutor, or crutch?: A grounded theory of cognitive scaffolding and offloading in AI-assisted programming education. International Journal of STEM Education. https://doi.org/10.1186/s40594-025-00592-w

q2 · i? · qualitative · r3

Comparative analysis of pre- and post-semester concept maps scored by two raters, triangulated with interview statements of dissonance between high project grades and perceived knowledge (e.g., fear of the AI-off final exam). The authors interpret these summaries "cautiously as converging, descriptive evidence rather than inferential tests".

"Figure 3 visually contrasts the average gains in concept map complexity, with the human-pair theoretical contrast group showing larger increases in cross-links and hierarchical organization."

Discussion¶

Related Claims¶

• Self-efficacy judgments correspond less well to behavior in learning settings than in performance settings — the general claim this finding bears on, with the studies behind it
• Self Assessment Improves Self Regulated Learning — related
• PAIR-C scaffolding shows mixed evidence for deep understanding and reduced misconceptions in emergent-phenomena instruction — related
• Educational cross-sectional studies link stronger AI reliance or dependence scores with lower academic self-efficacy, higher academic stress, and stronger performance expectations — related
• Process mining shows the ChatGPT group's revision loop centred on consulting ChatGPT, while the human-expert group strengthened connections between revising and reading, orientation and evaluation — related
• ESL students taught concept mapping show significantly greater gains in self-efficacy for learning from text than students taught an alternative learning strategy — related