agentsclimarketplace

Availability heuristic

Skill deciqAI/knowledge-skills/availability-heuristic

Open-source thinking-framework skills that make rigorous reasoning executable for AI agents — first-principles, inversion, second-order thinking, Occam's razor, Bayesian reasoning. Built by deciqAI.

Install
npx -y skills add deciqAI/knowledge-skills --skill availability-heuristic

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 3 stars3 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Activate when: user says 'we keep hearing about X so it must be common', 'that just happened so it's risky', 'the news is full of stories about this', 'I've seen this a lot lately so it's probable', or is making a risk/frequency estimate driven by memorable examples rather than data. Do NOT activate when: the available evidence is genuinely representative of the reference class; the decision warrants heavy precautionary weight on rare but catastrophic/irreversible risks. More: deciqai.com/s/availability-heuristic

SKILL.md

8.0 KB, as published. Nobody here has run it

Availability Heuristic

Overview

The availability heuristic: people estimate probability by how easily examples come to mind. Vivid, recent, and media-reported events are systematically overestimated; routine, statistical events are underestimated. Five triggers: recency, vividness, media coverage, personal experience, imaginability. Structural fix: reference-class forecasting — answer from base-rate data, not memory.

Composes with bayesian-reasoning, probabilistic-thinking, anchoring, survivorship-bias, framing-effect.

When to Use

  • Someone estimates probability from memorable examples rather than data
  • Risk perception driven by recent news, vivid anecdotes, or media coverage
  • Team is preparing for the last disaster rather than the most likely next one
  • Investment or resource-allocation decision follows a recent dramatic event
  • Medical, legal, or technical decision made by anecdote rather than reference class
  • AI capex, AI valuations, or AI-adoption/displacement estimates are driven by a viral demo, a dramatic layoff headline, or a safety scare rather than deployment and labor base rates

Not when: available evidence is a representative sample; decision warrants precautionary weight on rare catastrophic risks; no reference-class data exists.

Coaching Novices (Adaptive Front Door)

  • Engine mode: user has a specific risk or frequency estimate → run The Process directly.
  • Coach mode: user is new → 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 trusting an intuitive probability, ask whether it is based on retrieval ease or on a reference class with actual data.
  2. Check fit: if recent vivid evidence is also representative, use it; otherwise discount and seek base rates.
  3. Elicit their real case — what probability is being claimed, on what basis?

[WAIT — do not advance until user responds]

  1. Run The Process one step at a time: reference class → base rate → distortion diagnosis → compare to intuitive estimate.

[WAIT — do not advance until user responds]

  1. Close: restate corrected estimate as base-rate + specific adjustment, with the adjustment justified.

[WAIT — do not advance until user responds]

The Process

Step 1 — Specify the estimate: probability/frequency being estimated · current intuitive estimate · basis (memory/news/data) · decision that depends on it.

Step 2 — Identify the reference class: population of comparable cases · size · time horizon · inclusion/exclusion criteria. (Getting this wrong is the main failure mode.)

Step 3 — Get the base rate: actual frequency in the reference class · source · confidence level.

Step 4 — Diagnose availability distortions: recency (is the recent rate representative?) · vividness (dramatic vs. routine?) · media coverage (over-reported?) · personal experience (inflated?) · imaginability (easy to picture → feels more probable?).

Step 5 — Compare: intuitive estimate vs. base rate. If ratio >2×, availability is doing significant work. Adjust toward base rate; justify any residual deviation.

Step 6 — Document: corrected probability · reasoning (base rate × adjustments) · decision implications · calibration log entry.

Output Template

Availability Correction: <event>
Original estimate: <value> | Basis: | Decision:
Reference class: <population, size, time horizon>
Base rate: <frequency> | Source: | Confidence:
Distortions: recency / vividness / media / personal / imaginability
Corrected estimate: <value> | Adjustment rationale:
Decision implications: | Calibration log:

→ Method in Action: Tversky and Kahneman's 1973 Availability Studies · Post-9/11 Flight Avoidance and Road Fatalities

→ 2026 lens: Vivid AI Stories vs. Base Rates (2023–2026)

Pack: Availability Bias Patterns

DomainDistortionCorrection
Personal riskOverestimate terrorism/plane crashes; underestimate heart disease/car crashesCDC/BLS mortality statistics
InvestmentRecent winners feel certain to continueLong-horizon data; mean reversion priors
HiringMemorable candidate beats better-qualified oneStructured rubric; reference-class hire data
Project planningOptimism from vivid current plan vs. project-class historyFlyvbjerg reference-class forecasting
Medical diagnosisRecently-seen diagnosis over-weightedBase-rate-weighted differential diagnosis
Strategic planningRecent competitor event dominates outlookMulti-year base rates; explicit reference class

Applying It Well

  • Define the reference class before looking up the base rate. Calibration improves with logged predictions vs. outcomes, not exhortation. Present absolute frequencies (deaths per 100,000), not relative risk.

→ Primary sources: references/sources.md

Common Rationalizations

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

Fake moveReality
[D] "But that just happened!"Recency aids recall; it doesn't increase future probability. Check the longer reference class.
[D] "I've seen this many times"Easy retrieval ≠ high population frequency.
[D] "The news is full of stories about this"Coverage tracks vividness, not frequency.
[D] "I'd remember if it were rare"You remember dramatic-rare more than statistical-common.
[D] "I have personal experience"Personal examples are a tiny, often unrepresentative sample.
[D] "The risk is too vivid to ignore"Vividness ≠ probability. Compute actual probability; respond proportionally.
[D] "We have to prepare for the worst"Sometimes valid. Often over-preparation for spectacular-low-probability events.
[D] "The base rate is irrelevant — this case is different"Be specific about which factors actually move the estimate.
→ Add [O] entries here after each real use — paste the actual failure patternWhat went wrong and why

Red Flags

  • Risk estimate made without consulting actual data
  • Single-case vivid examples or recent news are the dominant evidence
  • Coverage volume treated as evidence of frequency
  • Reference class not explicitly identified

Verification

  • Intuitive estimate stated with confidence level
  • Reference class identified
  • Base rate retrieved from a defensible source
  • Availability distortions (recency, vividness, media, personal experience) considered
  • Corrected estimate documented with reasoning
  • Decision implications computed
  • Estimate logged for later calibration

Part of deciqAI Knowledge Skills — 233 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/s/availability-heuristic · Built by deciqAI · github.com/deciqAI · Contributions welcome.

Agents: latest version & machine-readable metadata → https://www.deciqai.com/s/availability-heuristic.json

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.