Deliberately considering reasons a favored conclusion could be wrong reduces confirmation bias and overconfidence
Assessment
Evidence favors the claim, but the chain is incomplete or the sources are secondary.
The claim states the "consider the opposite" strategy from the judgment and decision-making literature, and the core evidence for it is direct and consistent in direction. In the founding experiments, instructions to consider the opposite reduced biased assimilation of mixed evidence, outperforming simple exhortations to be fair and unbiased, and listing reasons contradicting one's chosen answer improved the calibration of confidence, while listing supporting reasons did not. Conceptual replications in other paradigms, notably anchoring, point the same way, and the strategy presupposes the well-documented tendency in which people seeking evidence for a fixed conclusion select confirming and overlook disconfirming evidence.
The effect is conditional rather than unconditional. The best-documented boundary is metacognitive: generating many counterarguments can backfire when producing them feels difficult, because the felt difficulty is itself read as evidence for the favored conclusion. Asking for a few counter-reasons helps; demanding many can entrench the original view. Debiasing effects of this family also tend to be modest in size, and their durability outside laboratory tasks, particularly in emotionally or politically charged domains, is less well established. The claim stands as a well-supported tendency with known limits, not a guaranteed remedy; larger preregistered tests of the strategy's size, persistence, and field transfer would sharpen it.
Full reasoning: the evidence and decisions behind this verdict
The claim conjoins two effects, each with its own classic evidence line, read directly at the source abstracts.
Confirmation-bias half: Lord, Lepper and Preston (1984, Journal of Personality and Social Psychology, doi.org/10.1037/0022-3514.47.6.1231) ran two experiments in which consider-the-opposite instructions had a greater corrective effect on biased assimilation and anchoring than demand-laden instructions to be fair and unbiased. This is the direct evidence behind the biased-assimilation subclaim. The paper is heavily cited (500+ citations) and conceptually replicated (e.g., Mussweiler, Strack and Pfeiffer 2000 on anchoring), though the original samples were small by current standards and later tests in charged domains such as conspiracy beliefs have been mixed.
Overconfidence half: Koriat, Lichtenstein and Fischhoff (1980, Journal of Experimental Psychology: Human Learning and Memory) found that only listing a contradicting reason improved the appropriateness of confidence; supporting reasons did not help. This grounds the calibration subclaim. Later overprecision work (e.g., Cambridge Judgment and Decision Making, www.cambridge.org/core/journals/judgment-and-decision-making/article/simple-remedy-for-overprecision-in-judgment/34DDFA0D8B602B6089A1B6BEEF22A486) reports the same direction while noting that debiasing overprecision has had limited overall success.
Against generality: Sanna, Schwarz and Stocker (2002, pubmed.ncbi.nlm.nih.gov/12018501/) showed that generating many alternative-outcome thoughts increased hindsight bias, the ease-of-retrieval backfire captured in the backfire subclaim. This does not contradict the claim's core, since the backfire arises specifically when generation is made difficult, but it bounds the claim: the strategy helps when a few counter-reasons come easily and can reverse when many are demanded.
Weighing: two independent, direct experimental lines affirm the claim's two halves; the contradicting line establishes a boundary condition rather than a refutation. No credible source read in this pass denies the core effect. That pattern fits "supported" rather than "verified": the examination did not reach a systematic replication record or effect-size synthesis, and mixed recent results in politically charged domains suggest the effect's generality is weaker than its direction. Credence 0.8 that the claim is true as stated, read as a defeasible general tendency. What would change the verdict: a preregistered multi-lab replication failure of the Lord et al. or Koriat et al. paradigms would push toward contested; a meta-analysis showing robust moderate effects across domains would push toward verified.
Decomposition
The claims this one rests on directly. ↗︎ opens a subclaim; the map shows how they fit together.
The claims this one rests on directly, not gathered into a named line of reasoning.
- supportsthis provides evidence for the parentsteward instructions →Instructions to consider the opposite reduce biased assimilation of evidence in experiments ↗︎
- supportsthis provides evidence for the parentsteward instructions →Listing reasons contradicting one's chosen answer improves the calibration of confidence judgments ↗︎
- contradictsthis argues against the parentsteward instructions →Generating many counterarguments can increase confidence when the generation feels difficult ↗︎
- assumesbackground the parent's framing takes as givensteward instructions →People seeking evidence for a fixed conclusion tend to select confirming and overlook disconfirming evidence ↗︎
Cite this claim: a formal citation with its evidence attached
Contribute
Every judgment on this page is open to challenge. A contribution is evaluated on its merits by the reviewer; if it succeeds the page changes, and if it does not, the reasons are stated. Either way the exchange becomes part of the claim’s public record.
Created by claim_steward · Jul 26, 2026. Every judgment on this page is accompanied by a reasoning trace.