Skip to the guide
About Free Assessments LifeGraph+

Behavior Lab · Applied guide

Treat a cognitive bias as a hypothesis about a decision—not a label for a person.

The useful question is not “Which biases do I have?” Everyone uses shortcuts. The useful question is whether a specific shortcut is distorting a specific choice, what evidence would reveal that, and which safeguard fits the method and context.

Evidence-boundedDecision trace · Source-method auditUpdated August 23, 2026
Short answer

Record the decision before you know the outcome. Capture your first estimate, reference class, alternatives, disconfirming evidence, incentives, and exit rule. Then review whether the process improved—without mistaking a good outcome for proof of unbiased reasoning.

Start narrowly

A bias name is not yet an explanation

Cognitive biases are systematic patterns that can pull judgment away from a relevant standard under particular conditions. The phrase does not mean that every shortcut is irrational, that careful people are exempt, or that an unfavorable result proves biased thinking. Time limits, missing information, incentives, expertise, emotion, and task design all matter.

Begin with one decision and one suspected mechanism. “Anchoring may be affecting my offer estimate because I saw the seller's number first” is inspectable. “I am an anchored thinker” is broad, identity-like, and hard to falsify. The first statement tells you to produce an independent estimate; the second mostly supplies a label.

Use a three-part claim. Name the cue, the proposed mechanism, and the observable consequence: “After cue X, I may be overweighting Y, which would show up as Z.” Then write what evidence would weaken the claim.

Source-method audit

Ask what produced the signal

Different methods reveal different slices of reasoning. Agreement across sources can strengthen a working hypothesis. Disagreement is not failure; it is information about context, self-perception, or task specificity.

What four evidence sources can and cannot tell you
SourceUseful forDoes not establish
Behavioral taskPerformance under a defined prompt, incentive, and scoring rule.A stable trait or the same pattern in every real-world domain.
Self-reportYour perceived tendencies, confidence, and recurring examples.Accurate recall, freedom from self-presentation, or actual task performance.
Prospective decision recordWhat you believed, forecast, and would accept before the outcome.That the recorded process captures every hidden influence.
Outcome or feedbackConsequences, calibration over repeated cases, and contextual learning.That one good result validates the process or one bad result invalidates it.

The LifeByLogic Cognitive Bias Test deliberately includes behavioral tasks and self-report. Compare those channels rather than collapsing them. Read the methodology and validation limits before using a result in an important decision.

Original utility

Build a prospective decision trace

Complete the trace before the outcome is known. Its purpose is to preserve the evidence state under which you made the choice, not to create paperwork for every minor preference.

  1. Decision and clock. State the choice, decision owner, deadline, and whether delay itself has a cost.
  2. Independent starting estimate. Before reviewing a proposed number or favored answer, write your own range and the assumptions behind it. If exposure already occurred, record that fact.
  3. Reference class or base rate. Identify a genuinely comparable set of past cases. Note where your case differs instead of forcing a false comparison.
  4. Live alternatives. Name at least two plausible options, including “do nothing for now” when real. Record why each could be right.
  5. Disconfirming evidence. Write the observation most likely to change your preference and how you will look for it.
  6. Incentives and identity. Note who benefits from each frame and which answer protects status, consistency, sunk effort, or a desired self-image.
  7. Exit rule. Decide what future cost, date, or signal would trigger reassessment. This limits sunk-cost reinterpretation later.
  8. Confidence range. Use a range, not a performance of certainty. Record what remains unknown and what information is not worth buying.

Copyable trace

Choice
I am deciding ___ by ___.
First estimate
Before outside anchors, my range is ___ because ___.
Reference class
Comparable cases are ___; my case differs because ___.
Best alternative
The strongest reason my leading option is wrong is ___.
Exit rule
I will reopen the decision if ___ happens by ___.

Pattern-to-safeguard map

Match the intervention to the suspected mechanism

Six common decision patterns and proportionate checks
Working patternObservable cueUseful check
AnchoringA first number or frame keeps pulling later estimates.Generate an independent range first; compare several anchors; expose assumptions.
Base-rate neglectA vivid case displaces relevant frequencies from comparable cases.Write the reference class and differences before discussing the unique story.
Confirmation searchYou ask only whether the favorite can work.Assign a disconfirming search, invite an independent critic, and specify change criteria.
Sunk-cost pullPast effort appears as a reason for future investment even when unrecoverable.Evaluate the next unit of cost from today; apply a prewritten exit rule.
Framing effectPreference changes when the same tradeoff is phrased as gain or loss.Restate the choice in absolute quantities, gains, losses, and status quo terms.
OverconfidenceRanges are too narrow or unknowns disappear from the record.Use intervals, track repeated forecasts, and compare confidence with calibration over time.

A checklist is not automatically debiasing. It must arrive before the relevant judgment, be easy enough to use, and target a known failure mode. In high-stakes professional settings, domain expertise and institutional design may matter more than a generic reminder.

Do not overclaim

Awareness, retention, and transfer are separate questions

Some debiasing interventions have reduced selected biases in experimental settings. That does not justify the claim that learning a list of biases makes someone broadly unbiased. An effect may be tied to a task, decay over time, or fail to transfer to a new context.

Lowest friction

Change the choice environment

Use defaults, independent estimates, delayed exposure to anchors, structured fields, or a second reviewer. The safeguard does not depend entirely on remembering a lesson at the right moment.

Reasoning prompt

Force an alternative model

Consider the opposite, run a premortem, or write the strongest competing explanation. The prompt should produce evidence, not ritual disagreement.

Repeated feedback

Track calibration

Record forecasts and confidence across comparable cases. Repetition helps distinguish a pattern from one memorable miss.

Transfer audit

Check the new context

Ask whether the task, incentives, expertise, time pressure, and feedback resemble the setting in which the intervention worked.

Prefer a safeguard that makes the desired process easier to execute. If a decision is consequential, seek relevant expertise, independent review, or formal procedures rather than relying on a self-assessment alone.

After the outcome

Audit process quality separately from luck

  1. Reconstruct from the frozen record. Do not rewrite what was knowable then using facts learned later.
  2. Score the process. Did you use a relevant base rate, seek disconfirming evidence, preserve alternatives, and follow the exit rule?
  3. Explain the outcome cautiously. Separate controllable decisions, uncontrollable events, measurement noise, and delayed effects.
  4. Update one safeguard. Keep, remove, or modify the intervention based on repeated usefulness—not embarrassment.
Four possible reviews: good process/good outcome; good process/bad outcome; poor process/good outcome; poor process/bad outcome. The third category is especially easy to miss because success can hide a fragile method.

Tool and evidence

Use the assessment as one source

Compare task performance with self-perception

The free Cognitive Bias Test combines seven behavioral tasks with twelve LifeByLogic-original self-report items. It is exploratory and educational. It does not diagnose a condition or prove how you will decide outside the task.

Take the free test Read the methodology

Sources and limits

  • Morewedge et al., 2015. Experimental tests of debiasing training for selected tasks. DOI record.
  • Retention and transfer review, 2021. Treats persistence and generalization as separate evidence questions. Open review.
  • Systematic review of professional cognitive biases, 2022. Documents heterogeneous contexts and methods. Open review.

This guide is educational. It does not establish that a named bias caused a decision, certify unbiased reasoning, or replace domain-specific review.