Brief: regression-to-the-mean
Research brief for regression-to-the-mean · how it was made.
Research brief - regression-to-the-mean
| Field | Value |
|---|---|
| Slug | regression-to-the-mean |
| Working title | Regression to the Mean |
| Article shape | Specimen Record |
| Category | Judgment Errors (je), next to Outcome Bias |
| Intent tags | decision-biases, at-work |
| AI authors credited | Grok 4.7 |
| Reviewer or checker | none |
| Date | 2026-09-28 |
1. Concept gate
Reader's real question:
- Why does a terrible week followed by a normal one look like proof that the scolding, the tool, or the remedy worked?
- Does this mean feedback and treatment do nothing?
Why this belongs in the catalog:
- The idea appears only as a side note inside false cause and Dunning-Kruger. The causal mistake deserves its own page.
- It is one of the most expensive untaught judgment errors: punishment looks effective and reward looks harmful because each is applied at an extreme.
Bug claim in one sentence:
The bug is regression to the mean, where a more ordinary result after an extreme one is credited to whatever happened in between, which makes the noise in the first score harder to see.
What this bug is not:
- Not the claim that coaching, medicine, or apologies never work. One rebound is a weak receipt.
- Not the gambler's fallacy. A coin has no usual level to return to.
- Not the hot-hand fallacy, which expects the extreme to continue.
- Galton's "mediocrity" is a statistical name for the ordinary level, not a moral verdict on people.
Nearby bugs to check:
| Nearby bug | Difference |
|---|---|
| Gambler's Fallacy | Chance is expected to balance. Regression is a noisy score moving toward a usual level. |
| Hot-Hand Fallacy | A streak is expected to continue. |
| False Cause | Any before-after story. Regression is the version where the first event is an extreme score. |
| Outcome Bias | Judges a decision by its result. Regression hands the sequel's credit to the intervention. |
| Illusory Correlation | Distinctive pairs stick. Interventions cluster where a rebound is already due. |
Concept gate: specific, distinct, grounded, and correctable - yes
2. Sources and research
| Source / concept | Year | What it supports | URL / DOI / note |
|---|---|---|---|
| Galton, F. "Regression towards mediocrity in hereditary stature." Journal of the Anthropological Institute of Great Britain and Ireland, 15, 246-263. | 1886 | Documents the inward pull in height and names it. Children of very tall parents tend to be tall and less extreme. | https://doi.org/10.2307/2841583 - verified 2026-09-28 |
| Tversky, A., & Kahneman, D. "Judgment under uncertainty: Heuristics and biases." Science, 185(4157), 1124-1131. | 1974 | Misconceptions of regression. Flight instructors praise a smooth landing and criticize a rough one; the next try moves inward; they conclude punishment works and reward backfires. The paper says that conclusion does not follow. | https://doi.org/10.1126/science.185.4157.1124 - bibliographic details checked against the 1974 Science text 2026-09-28 |
The article tells the instructor story in plain language and saves the names for Research, so Explained does not open with an author-year hook.
3. Article plan
Explained boundary:
- Extreme score = stable level + noise. The next score tends to be less extreme.
- The story in the gap is the bug.
- Real causes still exist. One before-and-after at an extreme is not the receipt.
Examples planned: praise backfires, yelling works, supplement, fund, employee of the month, software after the outage, apology.
Real-world scenarios: flight line, dashboard, slump, remedy at the peak.
Impact: punishment looks better than reward; programs get adopted or killed on a bounce; you stop learning what works.
Causes: extremes demand a story; you intervene at extremes on purpose, so you are often standing next to a rebound.
Prevention moves: ask what usually follows an extreme if you do nothing; compare with the person's usual level; precommit to what would count as more than a bounce; do not rewrite policy from one pair; list the other reasons a peak eases.
Questions & Answers: feedback and treatment; gambler's fallacy.
Reframing examples:
| Label | Original thought | Reframed thought |
|---|---|---|
| Praise backfires | The worse landing proves praise failed | Judge praise across many landings |
| The miracle month | The tool saved us | Calmer often comes after the worst month |
| The spoiled star | Employee of the month got lazy | An ordinary month after a peak is what noisy performance does |
Related thinking bugs: Gambler's Fallacy, Hot-Hand Fallacy, False Cause, Outcome Bias, Illusory Correlation.
5. Sign-off
- ☑ Belonging test passed
- ☑ Nearby bugs checked
- ☑ Sources verified or uncertainty labeled
- ☑ Examples are concrete
- ☑ Reframes are realistic, not forced optimism
- ☑ Credits model/person named