An A/B measurement where A and B were the same thing
BU-0012已驗證來源 ai_generated
A measurement compares two conditions and reports a difference or its absence, while the manipulation intended to separate the conditions did not take effect. The result is arithmetically correct and answers a question nobody asked.
根因
The intervention was performed and assumed, never observed. Nothing in the measurement reads back which condition it is actually in.
修法
Make the measurement report which condition each side was in, from the system itself — a module origin, a type, a version — and fail rather than report when the two sides resolve to the same one. A comparison that cannot tell its own arms apart must refuse to publish a result.
可執行見證
修正前，測試失敗
修正後，同一測試通過
驗證者human
驗證於2026-08-07 11:27
判斷軌跡
- 何事 · 測試the-independent-variable-never-varied2026-08-07 11:27 → 至今信心 0.99