Mental Models a field guide

Trap/Cognitive Bias/No. 0703

Outcome Bias

Outcome bias is the tendency to judge a decision by its result rather than the information available when it was made. Named by Jonathan Baron and John Hershey in 1988, it can make sound choices seem poor and reckless choices seem sound despite unchanged evidence.

a trap: easy to walk into
Evidence
Well established
Read
6 min
Links
10 connections
Useful when
Deciding under uncertainty · Fairness and ethics · Judging people and hiring · Money and investing · Risk and safety

01You've seen this when…

  1. in life

    You put your emergency savings into one speculative stock. Its price doubles, and you treat the gamble as proof of your investing skill.

  2. at work

    Two project leads follow the same launch checks. One launch hits an unrelated network outage; that lead receives the worse performance review.

  3. out in the world

    Officials evacuate a coastal neighborhood after a credible storm warning. The storm changes course, and residents condemn the evacuation as an obvious mistake.

02The idea

Once you know how a decision turned out, its quality can seem obvious. The successful investment was smart. The failed operation was a mistake. The evacuation that nobody needed was an overreaction.

Before the outcome was known, the person making the choice had incomplete information and faced several options, each with its own risks. Outcome bias means letting the ending change your judgment of the choice, even when the information available at the time stays the same.

A sound decision can produce a bad result. A reckless decision can get lucky. If a treatment has a 90% chance of success, a failure alone leaves the quality of the choice unresolved. To assess that choice, compare the available alternatives in light of the patient’s preferences and check whether the estimate was credible.

This differs from hindsight bias. Hindsight bias makes an outcome seem more predictable after it happens. Outcome bias makes the decision seem better or worse because of that outcome. You can admit that nobody could have predicted a failure and still unfairly downgrade the person who made the choice.

03Why it happens

  • The ending is easier to inspect than the reasoning. A profit, a death, or a missed deadline is concrete. Reconstructing the options and uncertainties takes work. The visible result becomes a shortcut for judging everything that preceded it.
  • The result supplies a convincing story. Once the outcome is known, earlier details can look like clues pointing toward it. A small concern becomes the warning everyone should have heeded; a lucky break becomes evidence of foresight.
  • Feelings spill into the verdict. Relief and gratitude can make the decision feel wise. Anger and grief can make it feel negligent. Those reactions are understandable. Judging what someone should reasonably have chosen beforehand requires examining the options and information available at the time.
  • Ordinary feedback encourages the shortcut. Repeating actions that succeed is often useful. But when outcomes contain substantial luck, that learning rule can reward bad reasoning and punish good reasoning. In safety settings, repeated lucky escapes can feed normalization of deviance: an unsafe practice starts to feel acceptable because nothing bad has happened yet.

04A worked example

In their 1988 research, Jonathan Baron and John Hershey asked participants to evaluate decisions described in written scenarios. Medical scenarios included a decision about surgery. Participants received information about the decision and its risks. The ending varied: the patient survived or died.

What it looks like The successful operation confirms good judgment, while the patient’s death exposes a poor decision. The ending seems to settle the evaluation.

What’s actually going on Participants rated decisions more favorably when the outcomes were favorable, even though the information supporting the choices was unchanged. The experiment separated decision quality from outcome by changing the ending without changing the grounds for the decision. The ratings nevertheless moved.

What would have helped Evaluating the recommendation before seeing the result, then discussing the result separately. Did the risks and the patient’s wishes make surgery preferable to the available alternatives? A death warrants investigation. Whether recommending surgery was reasonable depends on the grounds for that recommendation.

These were experimental scenarios, not evidence that any particular surgeon made a good or bad clinical decision.

05How to spot it

06What to do instead

  • Record the reasoning before the result arrives. A short decision journal should capture the options, assumptions, probabilities, and reasons for choosing. Otherwise, memory can quietly rewrite the original case.
  • Assess the choice before revealing the ending. When practical, give reviewers the information available at the time and ask for an initial assessment. Reveal the outcome afterward, in a separate learning discussion.
  • Ask what a reasonable person could have known. Distinguish warnings that were available and ignored from facts discovered only later. Judge the search for information as well as the final choice.
  • Review patterns, not just memorable cases. Across enough comparable decisions, check whether stated probabilities match observed frequencies. Calibration and a proper scoring rule help evaluate forecasts without treating every unsuccessful prediction as incompetence.
  • Separate decision review from harm response. First assess the choice and what needs to change. Separately address the damage, support affected people, and establish any responsibility. Compassion doesn’t require a distorted account of the decision.

07When it isn’t outcome bias

Outcomes contain information. A failed launch may reveal an omitted test. A treatment complication may expose a risk nobody had measured properly. Revising your view in response to newly discovered facts about the decision’s grounds is sensible Bayesian updating.

The distinction is whether you learned something about the decision’s grounds or merely reacted to its ending. Repeated failures deserve scrutiny, especially when comparable decisions succeed elsewhere. But even a track record needs context: someone handling the hardest cases may have worse results despite better judgment.

A sound decision depends on the quality of the process, even when someone follows a checklist. The process itself may be inadequate. A gamble with a positive average payoff may still be unacceptable if its downside threatens bankruptcy or someone’s life. Risk of ruin matters before the outcome arrives.

Finally, consequences can legitimately affect compensation, repair, and moral responsibility. Philosophers discuss this tension as moral luck. Outcome bias concerns the quality of a choice. The suffering caused by that choice remains relevant to how we respond.

08Roots

Jonathan Baron and John Hershey put readers in an uncomfortable position: judging a medical choice after learning whether the patient lived or died. They used written scenarios rather than operating rooms, which let them change the ending while keeping the decision’s background fixed. Their 1988 paper named the resulting distortion outcome bias.

There was already a neighboring problem. In 1975, Baruch Fischhoff had shown that knowing an outcome changes how predictable people think it was. Baron and Hershey focused on whether a decision deserved approval, separating that judgment from whether someone should have foreseen the ending. That distinction gives outcome bias its practical value in performance reviews, investment decisions, and accident investigations.

One possible adaptive explanation is that judging actions by their consequences offers a cheap learning rule. When an action reliably produces food, safety, or social approval, repeating it makes sense. This remains a plausible evolutionary hypothesis. The shortcut becomes misleading when luck, hidden causes, and rare events separate what someone chose from what eventually happened.

09How solid is this?

ContestedMixedUsefulEstablished

Controlled experiments show that changing an outcome can change evaluations of a decision whose supporting information stays fixed. The strength varies by task, and real outcomes can also reveal legitimate evidence about mistakes that were previously hidden.

10Connections

confused withcountered bycountered bycountered bycountered bycountered byleads toOutcome BiasHindsight BiasCounterfactualReasoningDecisionJournalCalibrationBayesianUpdatingNot written yetProperScoring RuleNormalizationof DevianceIllusionof ControlRisk of RuinNot written yetMoral Luck

11Origin and sources

Jonathan Baron and John Hershey named and experimentally documented outcome bias in decision evaluation in 1988.

  1. [1]Baron, J., & Hershey, J. C. (1988). Outcome bias in decision evaluation. Journal of Personality and Social Psychology, 54(4), 569–579.
  2. [2]Fischhoff, B. (1975). Hindsight ≠ foresight: The effect of outcome knowledge on judgment under uncertainty. Journal of Experimental Psychology: Human Perception and Performance, 1(3), 288–299.
  3. [3]Gneiting, T., & Raftery, A. E. (2007). Strictly Proper Scoring Rules, Prediction, and Estimation. Journal of the American Statistical Association, 102(477), 359–378.

Suggest an edit· Updated 2026-10-02