AI severs the correlation between artefact quality and cognitive engagement, breaking the artefact-as-proxy for learning¶

Claim · All claims
Evidence · 1 study · theoretical r3 · q1
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q1 i? The correlation between artefact quality and intellectual engagement has collapsed because the proxy was already vulnerable and AI removed the friction masking it. → Michael Rowe 2026

Evidence¶

Michael Rowe 2026¶

Michael Rowe. (2026). Problem-based learning and the structural conditions for productive AI integration. https://doi.org/10.35542/osf.io/haet3_v1

q1 · i? · theoretical · r3

Theoretical argument in a conceptual essay: assessment systems treated artefacts as inferential evidence of learning, and AI-generated polished outputs no longer require the engagement they were designed to evidence. The essay calls this "a Goodhart dynamic".

"AI breaks the artefact-as-proxy for learning. The correlation between artefact quality and intellectual engagement has collapsed, not because AI is powerful, but because the proxy was already vulnerable and AI removed the friction that was masking it."

Studies bearing on this claim¶

When generative AI helps make a learner's product, does the product's quality still show what the learner has learned and can do on their own? This table gathers the studies behind this claim and the 10 narrower claims that bear on it: 11 distinct studies (4 theoretical, 4 causal, 2 qualitative, 1 design), q1–q3, 3 of 11 with an effect size coded. One row per study, codes as each entry records them; no effect is averaged and no verdict is drawn. Built by scripts/synthesize_claim.py from eval/synthesis/; a row that looks wrong is fixed on the claim it comes from.

Discussion¶

Related Claims¶

• Engagement remains an observable and evidential proxy for learning that AI has not broken, though it is not invulnerable — related
• Course design created frictions: undeclared programming prerequisites, copied AI-generated tutorial solutions, intimidating blackboard turns, and AI-guessing cycles for students lacking evaluation knowledge — related
• Proposition 1: outcome-aligned ownership performances will predict later individual transfer beyond AI-assisted product quality alone — related
• Essay quality was highest for AI-generated drafts and lowest with no AI, negatively correlated with ownership, suggesting an ownership-quality tradeoff. — related
• ChatGPT-assisted students reported substantially lower ownership of their submitted code across self-attributed ownership, psychological ownership, and authorship attribution — related
• Participants report AI-generated database diagrams were less accurate than what students had previously produced, cautioning against assuming AI artifacts are better — a narrower finding that bears on this claim