Methodology
What this test measures, how it is scored, where the norm comes from, and what its limits are — all written here.
AIQ measures judgment about AI use, not memorizing tool names. Five dimensions, each with 4 scenario items plus 1 scale item:
| Code | Dimension | Measures |
|---|---|---|
| cog | Cognitive Understanding | Knowing what AI can and cannot do, and why it fails. |
| use | Tool Application | Breaking tasks down and choosing the right tools and approach. |
| ask | Prompting & Collaboration | Giving context, writing instructions, iterating across turns. |
| chk | Critical Verification | Checking sources, spotting hallucinations, knowing when not to trust. |
| eth | Ethics & Risk | Privacy, copyright, bias, and accountability. |
Scenario items: best option 2 points, acceptable 1, the rest 0 — so 'second-best but not crazy' and 'plain wrong' are distinguishable.
Scale items (1–5) are excluded from AIQ; they only feed the self-rating vs measured comparison on the result page.
Each dimension's raw score is converted to a 0–100 dimension score, then combined into the composite.
AIQ = 100 + 15 × z (z against the norm), clamped 40–160; norm baseline 100 with standard deviation 15, following the Wechsler tradition.
Scoring always runs on the server; the same answers always re-score identically; answer keys and option scores never reach the browser.
At launch we use a provisional norm (designer-estimated mean and standard deviation); its version number is shown on every result report.
Current version: zh-TW-prov-2026-10 (provisional).
After roughly 300 valid samples the norm is recalibrated and a new version marked; every attempt stores its raw answers, so a new norm can re-score everything.
After recalibration, past scores are re-scored under the new norm with old values kept — scores may move after a norm update; that is part of being honest.
Norm statistical parameters are server-side data and are not shown on the site.
This test is not a medical, psychological, or vocational assessment and must not be used as the sole basis for diagnosis, hiring, or any decision.
Each attempt reshuffles items from the 75-item bank, so two papers differ and score movement on retakes is normal.
Scores reflect AI judgment habits at the time of taking; AI tools evolve fast, so periodic retakes are recommended.
The gap between self-rating and measured performance is feedback in itself — a score is a starting point, not a ranking.