Lower engagement of cognitive control in children on ChatGPT
Primary school (grades 1–6) Adults / lifelong learning AI and learning Attention and memory Creativity
Links to claims
Limits (1)
| Claim | Basis |
|---|---|
| C2 | Small cross-sectional fMRI study: 6–7-year-olds using ChatGPT showed less integrated control/attention networks than adults using ChatGPT, with equal creativity. No no-AI condition and no learning outcome, so it neither supports nor tests C2 and cannot establish causation; context for the claim's unshown younger-pupil range. Abstract only. |
Review state: first-pass draft, pending human review.
Source note
Syntes: Horowitz-Kraus (2026): children chatting with ChatGPT in the scanner showed less integrated control and attention networks than adults, with equal creativity scores
Källa: Horowitz-Kraus, T.; Kenett, Y. N.; Link-Sourani, D.; Ashqar, N.; Farah, R. (2026). Lower engagement of cognitive control in children on ChatGPT. Cerebral Cortex 36(10), bhag167 (October 2026). https://doi.org/10.1093/cercor/bhag167
Verified content (abstract (Crossref record); full text not read, so sample sizes are not recorded here, checked 2026-10-13)
- Design (abstract): 6–7-year-old children and adults had the same naturalistic 5-minute co-creative dialogue with ChatGPT inside an fMRI scanner; functional connectivity computed within and between seven large-scale networks. Creativity assessed outside the scanner. Small sample (exact numbers not taken from the full text).
- Findings (abstract): creativity scores did not differ between groups. Adults showed stronger within-network connectivity in control and attention systems (most robustly frontoparietal) and more cross-network integration; children showed a less coherent and integrated control–attention–memory architecture.
- Limits: the comparison is children versus adults, both using ChatGPT; there is no condition without AI, so it cannot show that ChatGPT lowers children's cognitive control (the title's "lower" is relative to adults). Correlational brain measures; no learning outcome; small fMRI sample. It cannot establish causation.
Why it is in the library
Often cited in debates about children and AI. For this project it is context only: it says nothing yet about whether AI help lowers later learning.
Metadata as recorded
- ID
- AI-029
- Tier
- observational
- Closed-book follow-up
- not applicable (no learning outcome)
- Titel
- Lower engagement of cognitive control in children on ChatGPT
- Författare
- Horowitz-Kraus, T.; Kenett, Y. N.; Link-Sourani, D.; Ashqar, N.; Farah, R.
- År
- 2026 (Cerebral Cortex 36(10), bhag167 (October 2026))
- Källa/DOI
- https://doi.org/10.1093/cercor/bhag167
- URL
- https://doi.org/10.1093/cercor/bhag167
- Typ
- cross-sectional fMRI comparison of 6–7-year-old children and adults during a 5-minute co-creative dialogue with ChatGPT (small sample)
- Åldersgrupp
- grundskola-lag;vuxen-livslangt
- Teman
- ai-och-larande;uppmarksamhet-och-minne;kreativitet
- Hämtad
- 2026-10-13
- Read
- abstract (Crossref record); full text not read, so sample sizes are not recorded here
- Tillagd
- suggested by Jonatan (via Noll) 2026-10-13; links pending human review
The original files (PDF or abstract) are kept in the project archive and are not published here.