Claim	Basis
K19 (candidate)	Note carries the bank kill sentence behind K19 (group K19); no change in review

Claim

Basis

Note carries the bank kill sentence behind K19 (group K19); no change in review

Claim	Basis
C2	Higher-education meta g ≈ 0.50; not strictly closed-book. Open test for C2.

Claim

Basis

Higher-education meta g ≈ 0.50; not strictly closed-book. Open test for C2.

Source note (Swedish, verbatim from the Kunskapsbank)

Syntes: Fan, Ke, Chen & Lv (2026) — GenAI och läranderesultat i högre utbildning

ID: AI-002
Tier: meta
Closed-book follow-up: nej / sällan strikt (författarna varnar: uppgiftsprestation ≠ hållbart lärande)
HE vs K–12: HE only
Novelty decay: oklar; många korta interventioner
Utfallslabel: prestation, inte lärande

Metadata

ID: AI-002
Tier: meta
N: 36 studier, 132 ES, N=7 229
Population: vuxen / HE
Country: blandad (eng/kinesiska inkl.)
Subject: GenAI learning outcomes
Duration: varierar, ofta kort

Finding

Hedges’ g ≈ 0,50 overall i HE. Starkast collaborative/blended. Utan strikt closed-book: tolka som prestation, inte lärande.

Mechanism

Pedagogisk inbäddning (collaborative/blended) modererar mer än modellversion; risk för cognitive offloading.

Limits

Ingen K–12; closed-book sällan; heterogenitet.

This claim dies if …

K–12-metas med obligatorisk closed-book follow-up visar g≈0 eller negativ retention trots positiv practice-performance.

Swedish implication

Measure practice-vs-closed-book-gap innan GenAI skalas; stop att importera HE g≈0,5 som K–12-lärandeeffekt.

Exploring the effect of GenAI on learning outcomes in higher education: a three-level meta-analysis

Links to claims

Supports (1)

Challenges (1)