clawhub

Hindsight Bias

Item: Hindsight Bias
Rating: 8.3
Author: Implexa

Activate when: someone says 'I knew it all along' or 'we should have seen it coming'; a post-mortem is blaming someone for not predicting an outcome; a team...

view source

installs

306

stars

karma

SkillRank score ↗

8.3/ 10

evaluated by implexa, claude-haiku-4-5 · 2026-07-30

hindsight-bias detects post-outcome blame and memory distortion by reconstructing pre-outcome information sets and decomposing three distinct bias components. it installs pre-commitment documentation and blind post-mortems as structural fixes rather than relying on vigilance alone.

structure

9.0

trigger phrases

9.0

procedure

9.0

edge cases

7.0

documentation

8.0

view original SKILL.md from clawhubclick to expand

---
name: hindsight-bias
description: "Activate when: someone says 'I knew it all along' or 'we should have seen it coming'; a post-mortem is blaming someone for not predicting an outcome; a team is reviewing a past decision and the outcome is coloring the judgment; a decision-maker is being evaluated on what happened rather than what was knowable at the time.
  Do NOT activate when: contemporaneous pre-decision records exist and match current memory (bias is bounded); the goal is explicitly pattern-recognition from outcomes rather than evaluating the original decision-maker. More: deciqai.com/c/hindsight-bias"
---

# Hindsight Bias

## Overview

**Hindsight bias** — the "I knew it all along" effect — is the tendency, *after* learning an outcome, to misremember your prior judgment as having been closer to that outcome than it actually was. Fischhoff (1975) demonstrated three distinct components: **memory distortion** ("I said it would happen"), **inevitability** ("it had to happen"), and **foreseeability** ("anyone should have seen it"). Each requires a different countermeasure; the structural fix for all three is pre-commitment documentation.

Composes with `probabilistic-thinking`, `premortem`, `confirmation-bias`, and `survivorship-bias`.

## When to Use

- A post-mortem is becoming an exercise in blame for an "obviously foreseeable" outcome
- Someone says "I knew this would happen" without contemporaneous records
- A decision is being evaluated against its outcome rather than the information available when made
- An investor, judge, or jury evaluates a past decision with knowledge of how it turned out
- Someone calls a market move "obvious in retrospect" — e.g. the AI boom, AI capex buildout, or frontier-AI valuations were "clearly inevitable" / "anyone should have seen the AI adoption wave coming"

**Not when:** pre-commitment documentation exists and matches current memory; the goal is pattern-recognition from outcomes; the failure to predict was a genuine process failure with clear pre-outcome signals.

## Coaching Novices (Adaptive Front Door)

- **Engine mode:** user has a specific post-outcome narrative → run The Process directly.
- **Coach mode:** user is unfamiliar or has no concrete case → guide step by step.

In Coach mode, respond one step at a time. Each [WAIT] is a hard stop — output only that step's question, then stop.

1. One-line: before judging a past decision by its outcome, ask what was actually knowable in advance — check the contemporaneous record, not your current memory.
2. Check fit against When to Use / When NOT to use. If records exist and match memory, bias is bounded.
3. Elicit the specific outcome and the claim of foreseeability. What outcome happened? Who is being credited/blamed?
> **[WAIT — do not advance until user responds]**
4. Run The Process one step at a time: reconstruct the pre-outcome information set, decompose the three components, evaluate decision process vs. outcome.
> **[WAIT — do not advance until user responds]**
5. Close by naming the insight uncovered and the structural fix (decision journal, pre-mortem, blind post-mortem).
> **[WAIT — do not advance until user responds]**

## The Process

**Step 1 — State the outcome and the claim:** outcome that occurred / who is being credited-blamed / claim of foreseeability (verbatim) / time elapsed.

**Step 2 — Reconstruct the pre-outcome information set:** what was knowable / what was genuinely uncertain / what contemporaneous records exist / consensus view at the time. *Nixon test: if informed contemporaries gave the outcome ≤30%, "anyone should have seen it" is hindsight bias.*

**Step 3 — Decompose the three components:** memory distortion (current memory vs. records) / inevitability (is outcome framed as the only possible result?) / foreseeability ("anyone should have seen it" applied to genuinely uncertain ex-ante info?). Countermeasures: memory → pull records; inevitability → list 3-5 alternative outcomes; foreseeability → identify what pre-outcome info would have made it predictable.

**Step 4 — Evaluate the decision process, not the outcome:** was the decision reasonable given available info / expected value across plausible outcomes / would you make the same decision again with the same info. *Bridgewater matrix: good decision + bad outcome = bad luck (don't change). Bad decision + good outcome = good luck (don't celebrate).*

**Step 5 — Install pre-commitment documentation:** decision journal (who/what/when/why/probability) / pre-mortem record / public prediction logged to internal forum / pre-registered evaluation criteria.

**Step 6 — Run a blind post-mortem:** read decision journal before outcome data / evaluate against recorded reasoning / introduce outcome data only then / distinguish process error from outcome variance.

## Output Template

```markdown
# Hindsight Bias Analysis: <outcome>
Outcome / Foreseeability claim (verbatim) / Who is credited-blamed:
Pre-outcome info set (knowable / uncertain / records / consensus / ex-ante probability):
Three-component diagnosis (memory distortion Y/N / inevitability Y/N / foreseeability overclaim Y/N):
Decision-process evaluation (process quality 1-5 / reasonable Y/N / EV / same decision again Y/N):
Structural fix (decision journal owner / pre-mortem owner / blind post-mortem protocol Y/N):
```

*→ Method in Action: [Baruch Fischhoff's Nixon-China Trip Study, 1972-1975](examples/baruch-fischhoffs-nixon-china-trip-study-1972-1975.md) · [Anesthesia Malpractice Outcome-Bias Study, 1991](examples/caplan-posner-cheney-anesthesia-outcome-study-1991.md)*
*→ 2026 lens: ["The AI Boom Was Obviously Coming" — retrospective inevitability in the ChatGPT/Nvidia era (2022–2026)](examples/ai-boom-obviously-coming-hindsight-2022-2026.md)*

## Pack: Hindsight Bias Patterns

| Domain | Common hindsight manifestation | Structural countermeasure |
|---|---|---|
| Investment | "Obvious in retrospect" trade | Decision journal with logged thesis and probability |
| Engineering | "How did anyone miss that bug?" | Blameless post-mortem; reconstruct info state at time |
| Hiring | "I knew this person wouldn't work out" | Logged interview rubric and pre-hire confidence rating |
| Legal | "The defendant should have foreseen the harm" | Ex ante risk assessment; "reasonable person at the time" |

## Applying It Well

- Pull contemporaneous records before forming any judgment about what was "knowable"
- Evaluate decision quality and outcome quality on separate axes — never conflate them
- The structural fix is records, not vigilance — warnings reduce but do not eliminate the bias

*→ Primary sources: [references/sources.md](references/sources.md)*

## Common Rationalizations

**[D] = designed upfront | [O] = observed in real use. [O] entries are more valuable.**

| Fake move | Reality |
|---|---|
| [D] "I remember exactly what I said at the time" | Confident-memory subjects were *just as biased*. Memory without records is unreliable. |
| [D] "But it was so obvious in retrospect" | Hindsight makes everything look obvious. Test: did informed contemporaries give it >50% probability before the outcome? |
| [D] "Anyone with sense would have predicted this" | Check the contemporaneous record — pre-event surveys, market prices, expert forecasts. |
| [D] "The outcome was inevitable given the structure" | List 3-5 plausible alternative outcomes the same structure could have produced. |
| [D] "Process doesn't matter; outcome is what matters" | Outcome alone confounds skill and luck. Evaluate them separately. |
| [D] "We learned the lesson — that's all that matters" | If the "lesson" is hindsight-biased, the next decision will be miscalibrated. |
| *→ Add [O] entries here after each real use — paste the actual failure pattern* | *What went wrong and why* |

## Red Flags

- Post-mortem conducted with no pre-mortem record to compare against
- "Obvious in retrospect," "anyone should have seen it," "we knew this" without contemporaneous documentation
- "Lesson learned" focuses on the specific failure mode, not on the broader uncertainty space
- Decision-maker blamed/credited based primarily on outcome rather than process
- Emotional or financial stakes are high — the bias is *larger* in high-stakes cases

## Verification

- [ ] Contemporaneous records pulled (emails, memos, forecasts, market prices, decision journals)
- [ ] Three components diagnosed separately (memory / inevitability / foreseeability)
- [ ] At least 3 plausible alternative outcomes enumerated for the same starting structure
- [ ] Decision quality evaluated separately from outcome quality
- [ ] Ex-ante probability estimated using pre-outcome information only
- [ ] Pre-commitment documentation installed for future similar decisions
- [ ] Post-mortem protocol blind to outcome data until after process review

---

*Part of **deciqAI Knowledge Skills** — 227 open-source thinking skills that make rigor executable for AI agents. The same skills power every deciqAI agent, which runs them autonomously to operate your company. **See it run → https://www.deciqai.com/c/hindsight-bias** · ⭐ Star the repo → https://github.com/deciqAI/knowledge-skills · Contributions welcome.*

*Agents: latest version & machine-readable metadata → https://www.deciqai.com/s/hindsight-bias.json*

related skills

semantically similar in the cross-vendor index

clawhub

81% match

Narrative Fallacy

Activate when: user is reading a business case study or founder biography and drawing lessons; a post-mortem produced a neat single root cause; someone says...

don't have the plugin yet? install it then click "run inline in claude" again.

added explicit inputs (contemporaneous records, context requirements, mode selection), structured procedure with pre-flight checks and coach/engine mode branching, comprehensive decision points for activation and edge cases, detailed output-contract template with markdown structure, and outcome signals for success verification across both modes.

Hindsight Bias

Intent

Hindsight bias , the "I knew it all along" effect , distorts how we remember past decisions and judgments after learning their outcomes. It manifests in three distinct ways: memory distortion (misremembering what you said), inevitability framing (treating outcomes as the only possible result), and foreseeability overclaim (claiming "anyone should have seen it" when the outcome was genuinely uncertain ex-ante). use this skill when a post-mortem is turning into blame for an "obviously foreseeable" outcome, when someone claims they predicted something without records, or when a decision is being evaluated against its outcome rather than the information available when it was made. the structural fix for all three components is pre-commitment documentation (decision journals, pre-mortems, blind post-mortems, logged predictions).

Inputs

External connections / required documentation:

contemporaneous records: decision memos, emails, meeting notes, forecasts, market prices, interview rubrics, risk assessments, or decision journals from the time the decision was made (not reconstructed after the outcome)
pre-mortem records (if any exist): documented alternative outcomes or failure modes considered before execution
consensus view at the time: expert forecasts, market prices, survey data, or informed contemporaries' stated beliefs pre-outcome
outcome data: what actually happened, timing, and who is attributing credit/blame

Context needed:

specific outcome that occurred (concrete, measurable)
verbatim claim of foreseeability ("I knew it would happen," "we should have seen it coming," "obviously inevitable")
who is being credited or blamed
time elapsed between decision and outcome
decision domain (investment, engineering, hiring, legal, operations, etc.)
stakes of the evaluation (whether reputational, financial, or career impact is high)

Mode selection:

engine mode: user has a specific post-outcome narrative and contemporaneous context; run The Process directly
coach mode: user is unfamiliar with the skill, has no concrete case, or is learning hindsight-bias recognition; guide step by step with explicit waits

Procedure

Pre-flight check (before starting The Process):

confirm the skill applies: the outcome has occurred, a claim of foreseeability or inevitability has been made (explicit or implicit), and the goal is evaluating a past decision or decision-maker. if contemporaneous records exist and match current memory, bias is bounded; note this and exit early.
identify mode: does the user have a specific post-outcome narrative with contemporaneous context (engine mode), or should you guide step by step (coach mode)? if coach mode, move to coach sequence below. if engine mode, proceed to The Process.

Coach Mode (Adaptive Front Door) , one step at a time, hard stop at each [WAIT]:

output the one-line frame: "before judging a past decision by its outcome, ask what was actually knowable in advance , check the contemporaneous record, not your current memory." then [WAIT , do not advance].
when user responds, elicit the specific outcome and foreseeability claim. ask: what outcome happened? who is being credited or blamed? what is the exact claim of foreseeability or inevitability? then [WAIT].
when user responds, begin The Process (step 1 below), but run it one step at a time. after each step, [WAIT , do not advance until user responds].
after completing all five steps of The Process, close by naming the insight uncovered and the structural fix (decision journal, pre-mortem, blind post-mortem, logged predictions). then [WAIT].

Engine Mode , The Process (run all steps in sequence):

step 1 , state the outcome and the claim: output the specific outcome that occurred, who is being credited or blamed, the verbatim claim of foreseeability (e.g., "I knew it all along," "it was obviously coming," "anyone should have seen it"), and time elapsed since the decision. confirm domain (investment, hiring, engineering, legal, etc.).
step 2 , reconstruct the pre-outcome information set: identify what was actually knowable at the time (expert forecasts, market prices, contemporaneous records, consensus view). identify what was genuinely uncertain (alternative outcomes with non-trivial probability). pull and reference contemporaneous records. apply the nixon test: if informed contemporaries gave the outcome ≤30% probability pre-outcome, "anyone should have seen it" is hindsight bias. output the ex-ante probability estimate based on pre-outcome information only.
step 3 , decompose the three components: evaluate each separately:
- memory distortion: compare current memory of what the person said/believed vs. what contemporaneous records show. if records don't exist, mark as "no baseline."
- inevitability: is the outcome being framed as the only possible result? list 3-5 plausible alternative outcomes the same structure could have produced.
- foreseeability overclaim: was "anyone should have seen it" applied to genuinely uncertain ex-ante information? check: what pre-outcome information would have made it predictable? does contemporaneous data show that information was available? output a diagnosis for each: Y/N + evidence.
step 4 , evaluate the decision process, not the outcome: separate decision quality from outcome quality (bridgewater matrix: good decision + bad outcome = bad luck; bad decision + good outcome = good luck). ask: was the decision reasonable given the available information at the time? what was the expected value across plausible outcomes? would you make the same decision again with the same information and uncertainty? output a process-quality score (1-5) and Y/N on whether the decision was reasonable ex-ante.
step 5 , install pre-commitment documentation: prescribe one or more structural fixes:
- decision journal: logged entry with who/what/when/why/probability, owned by the decision-maker
- pre-mortem record: documented alternative outcomes and failure modes before execution
- public prediction logged to internal forum: timestamp and prediction before outcome
- pre-registered evaluation criteria: what counts as success/failure before the outcome is known
- blind post-mortem protocol: review decision journal and recorded reasoning before introducing outcome data output a recommendation for which fix(es) fit the domain and stakes.
output the full analysis using the Output Contract template (step 6, below).

Decision Points

Activate hindsight bias skill if:

someone says "I knew it all along," "we should have seen it coming," or "obviously inevitable" without contemporaneous records to back it up
a post-mortem is framing a past decision as obviously wrong in hindsight
a decision-maker is being blamed or credited primarily based on outcome, not process
a team is reviewing a past decision and the outcome is visibly coloring judgment
stakes are high (financial, reputational, or career impact) , the bias is larger in high-stakes cases

Do NOT activate if:

contemporaneous pre-decision records exist and match current memory (the bias is bounded; note this and move on)
the explicit goal is pattern-recognition from outcomes rather than evaluating the original decision-maker's judgment
the failure to predict was a genuine process failure with clear pre-outcome signals (e.g., a bug was flagged in a code review and ignored; a risk was logged and never escalated)

Coach mode triggers:

user says "I'm not sure how to apply this" or has no concrete case
user is unfamiliar with hindsight bias and needs scaffolding
stakes are high and the user needs to slow down and reason step-by-step

Engine mode triggers:

user presents a specific post-outcome narrative with contemporaneous context
user wants a rapid hindsight-bias diagnosis
user is experienced with the skill and wants to run The Process directly

Edge cases / handling:

no contemporaneous records exist: mark memory distortion as "no baseline" and focus on inevitability and foreseeability components. recommend installing decision journals immediately.
outcome was genuinely unpredictable (ex-ante probability <10%): the bias is still present (memory distortion, inevitability framing), but foreseeability overclaim is less severe. surface the distinction clearly.
high emotional or financial stakes: bias is larger; double-check for inevitability framing and memory distortion. slow down and pull records obsessively.
contemporaneous records are partial or ambiguous: flag the gap and anchor on what is explicit in the record, not what is implied or reconstructed.
multiple decision-makers: identify separately which components apply to each person. don't conflate collective hindsight with individual bias.

Output Contract

output the full analysis in the following markdown format:

# Hindsight Bias Analysis: <outcome>

## Outcome & Foreseeability Claim
- outcome that occurred: <specific, measurable>
- who is credited/blamed: <name or role>
- verbatim claim: <exact quote or paraphrase>
- time elapsed: <duration between decision and outcome>
- domain: <investment / engineering / hiring / legal / operations / other>

## Pre-Outcome Information Set
- knowable at the time: <what was documented, observable, or expert consensus>
- genuinely uncertain: <alternative outcomes with non-trivial probability>
- contemporaneous records: <list sources: decision memo, forecast, market price, email, survey, etc.>
- ex-ante probability (pre-outcome): <% or range, based on pre-outcome info only>
- nixon test result: <did informed contemporaries give outcome ≤30% probability? Y/N>

## Three-Component Diagnosis
- **memory distortion:** <Y/N + evidence: current memory vs. records, or "no baseline">
- **inevitability framing:** <Y/N + list 3-5 alternative outcomes that same structure could have produced>
- **foreseeability overclaim:** <Y/N + check: was "anyone should have seen it" applied to genuinely uncertain ex-ante info? what pre-outcome info would have made it predictable?>

## Decision-Process Evaluation
- process quality (1-5 scale): <score>
- was the decision reasonable given available information: <Y/N + reasoning>
- expected value across plausible outcomes: <estimate or qualitative assessment>
- would you make the same decision again with the same information: <Y/N>
- bridgewater diagnosis: <good/bad decision + good/bad outcome = luck/skill assessment>

## Structural Fix (Pre-Commitment Documentation)
- recommended fix(es): <decision journal / pre-mortem record / public prediction / blind post-mortem / pre-registered criteria>
- owner: <who installs and maintains>
- timeline: <when to install>

## Insight & Next Step
- hindsight-bias component uncovered: <which of the three dominated this case>
- action: <decision-maker, team, or org to execute>

data format: markdown, plain text, or JSON (for machine-readable intake). file location: output to the user's chat/message window, or saved to decision-journal repo under hindsight-bias-analyses/<decision-name>.md.

Outcome Signal

the user knows the skill worked when:

they can name the specific hindsight-bias component that applied (memory distortion, inevitability framing, or foreseeability overclaim) with evidence
they have reconstructed the pre-outcome information set and can articulate what was actually knowable vs. genuinely uncertain at the time
they can separate decision quality from outcome quality and explain whether the decision was reasonable ex-ante (separately from how it turned out)
they have identified a structural fix (decision journal, pre-mortem, blind post-mortem, or logged prediction) and assigned an owner
they can articulate the distinction between "we made a bad decision and got lucky" vs. "we made a good decision and got unlucky" , or vice versa , instead of conflating outcome with quality
in coach mode: they can walk through The Process without guidance and apply it to a new case
in engine mode: they have a complete, filled-in output-contract template with all five steps resolved

credits: original skill design by deciqAI; methodology grounded in Baruch Fischhoff's hindsight-bias research (1975-); examples drawn from anesthesia malpractice outcome-bias study (Caplan et al., 1991) and 2022-2026 AI-boom retrospective analysis. part of deciqAI knowledge-skills open-source repo. see deciqai.com/c/hindsight-bias for examples and live agent execution.