AI makes you smarter but none the wiser: The disconnect between performance and metacognition
Adults / lifelong learning AI and learning Metacognition and self-regulation
Links to claims
Supports (1)
| Claim | Basis |
|---|---|
| K52 (candidate) | By analogy: two studies (N = 246, 452) of adults solving LSAT problems with AI: performance up about 3 points vs a norm population but self-estimates about 4 points too high; higher AI literacy, lower accuracy. Performance with AI, not unaided learning; no randomised no-AI group. Abstract only. |
Review state: first-pass draft, pending human review.
Source note
Syntes: Fernandes et al. (2026) — people solving reasoning problems with AI scored better but overestimated themselves even more
Källa: Fernandes, D., Villa, S., Nicholls, S., Haavisto, O., Buschek, D., Schmidt, A., Kosch, T., Shen, C. & Welsch, R. (2026). AI makes you smarter but none the wiser: The disconnect between performance and metacognition. Computers in Human Behavior 175, 108779 (February 2026). https://doi.org/10.1016/j.chb.2025.108779
Verified content
- [abstract] Study 1: 246 adults used AI on 20 LSAT logical-reasoning problems; performance was 3 points above a norm population, but they overestimated their performance by 4 points.
- [abstract] Higher AI literacy went with lower metacognitive accuracy.
- [abstract] The usual Dunning-Kruger pattern on this task disappeared with AI use.
- [abstract] Study 2 (N = 452) replicated the findings.
Limits
- [assessment] Abstract only. Performance with AI, not learning or later unaided performance; compared with a norm population, not a randomised no-AI group. Adults, reasoning puzzles.
Why it is in the library
A large, replicated human-AI study of exactly the gap K52 is about (felt vs real performance), and a warning that AI-savvy users judge themselves worst.
Added 2026-11-01 by the project (weekly source sweep, kind: adjacent (human-AI interaction); read: abstract). [abstract] = from the published abstract; [full text] = read in the full text; [assessment] = the project's own assessment.
Metadata as recorded
- ID
- AI-046
- Tier
- experiment
- Titel
- AI makes you smarter but none the wiser: The disconnect between performance and metacognition
- Författare
- Fernandes, D., Villa, S., Nicholls, S., Haavisto, O., Buschek, D., Schmidt, A., Kosch, T., Shen, C. & Welsch, R.
- År
- 2026 (Computers in Human Behavior 175, 108779 (February 2026))
- Källa/DOI
- https://doi.org/10.1016/j.chb.2025.108779
- URL
- https://doi.org/10.1016/j.chb.2025.108779 (listed as open access)
- Typ
- two online studies (N = 246 and N = 452) of adults solving LSAT logical-reasoning problems with AI, compared with a norm population
- Åldersgrupp
- vuxen-livslangt
- Teman
- ai-och-larande;metakognition-och-sjalvreglering
- Hämtad
- 2026-11-01
- Paywall
- open access (per OpenAlex)
- Tillagd
- source sweep 2026-11-01 (kind: adjacent (human-AI interaction)); links pending human review
The original files (PDF or abstract) are kept in the project archive and are not published here.