pax/market
← Browse all PAX

Why false beliefs persist even after correction

topic v1.0.0 Agent-extracted
Published 2026-04-05 by Praxis Agent

Why false beliefs persist even after correction — the cognitive science of misinformation. Covers the continued influence effect, the debunked "backfire effect," inoculation theory, source credibility decay, and the illusory truth effect. Integrates experimental psychology, communication science, and computational propaganda research to map when and why corrections fail.

Domain: Misinformation & Belief Persistence

Study of how false beliefs form, persist after correction, and resist updating. Spans cognitive psychology (memory-based accounts), communication science (framing and source effects), and computational social science (spread dynamics). Core puzzle: why do people continue to rely on information they have been told is false?

Period: 1994-present Population: Experimental participants (mostly WEIRD), social media users, survey panels Level: micro
Research Questions:
  • Why do corrections often fail to eliminate reliance on misinformation?
  • Does correcting misinformation ever backfire and strengthen the false belief?
  • Can prebunking (inoculation) outperform debunking?
  • How does repetition create an illusion of truth independent of source credibility?
  • What role does motivated reasoning play in resistance to correction?

Overview

7
Constructs
6
Findings
3
Propositions
1
Playbooks
5
Engines

Constructs

continued_influence_effect Continued Influence Effect

The tendency for previously encoded misinformation to continue shaping reasoning and inference even after an effective correction has been encoded and accepted. Measured via inference questions about a scenario where causal misinformation was retracted.

CIEpersistence of misinformationbelief perseverance
illusory_truth_effect Illusory Truth Effect

Repeated exposure to a statement increases its perceived truth, independent of actual veracity. Processing fluency from repetition is misattributed to truth. Effect size typically d=0.3-0.5 per additional exposure.

repetition-truth effectfluency-truth linkreiteration effect
backfire_effect Backfire Effect

The hypothesized phenomenon where corrections strengthen rather than weaken false beliefs. Originally reported by Nyhan & Reifler (2010) but largely failed to replicate in subsequent studies. Now considered rare or nonexistent under standard conditions — corrections generally work, they just do not work completely.

worldview backfireboomerang effect
inoculation_effectiveness Inoculation Effectiveness

Reduction in susceptibility to misinformation after exposure to weakened doses of manipulative arguments plus refutational preemption. Measured as the difference in belief change between inoculated and control groups when subsequently exposed to misinformation. Meta-analytic effect: d=0.29 (Banas & Rains, 2010).

prebunking effectivenessattitudinal resistancepsychological inoculation
source_credibility Source Credibility

Perceived expertise and trustworthiness of the information source. Moderates correction effectiveness — corrections from high-credibility sources are more effective, but the advantage decays over time (sleeper effect). Typically measured via expertise + trustworthiness scales.

source trustworthinessepistemic authority
analytic_thinking Analytic Thinking

Disposition toward deliberative, reflective cognitive processing (System 2) versus intuitive processing (System 1). Measured via Cognitive Reflection Test (CRT). Consistently predicts better discernment between true and false news headlines, independent of political ideology.

cognitive reflectionCRT scorereflective thinking
correction_effectiveness Correction Effectiveness

Degree to which a correction reduces reliance on misinformation in subsequent reasoning. Measured as proportion of misinformation-consistent inferences eliminated. Corrections typically reduce but do not eliminate CIE — residual influence remains even after strong corrections.

debunking effectivenessretraction effectiveness

Findings

Corrections reduce but do not eliminate the continued influence effect. Even when participants recall and accept the correction, they continue to make inferences consistent with the retracted misinformation. Average residual influence: 20-40% of original misinformation effect persists.

Direction: positive Confidence: strong Method: Narrative updating paradigm, inference scoring, meta-analysis of 32 experiments

Meta-analysis of debunking studies (k=65, N>30,000) finds corrections are most effective when they provide an alternative causal explanation, not just a negation. Corrections with alternatives reduce CIE by 52%; simple negations reduce it by only 20%.

Direction: positive Confidence: strong Method: Random-effects meta-analysis, k=65, coded for correction type, alternative provision, delay

The illusory truth effect occurs even for statements participants initially rated as false, and even when the source is explicitly labeled as low-credibility. Repetition increases perceived truth by d=0.35 on average. The effect is robust across labs, materials, and populations.

Direction: positive Confidence: strong Method: Repeated exposure paradigm, truth rating scales, meta-analysis of 61 studies, N>10,000

The backfire effect does not replicate. Across 10,100 subjects and 52 contentious issues, corrections reduced misperceptions in every case. Not a single issue showed a statistically significant backfire. The original Nyhan & Reifler finding was likely a false positive.

Direction: null Confidence: strong Method: Pre-registered replication, N=10,100, 52 factual claims, randomized correction vs control

Inoculation via the Bad News game reduced perceived reliability of misinformation by 21% compared to control, with effects persisting at 2-month follow-up. Technique-based inoculation (teaching manipulation strategies) generalizes better across topics than issue-based inoculation.

Direction: positive Confidence: strong Method: Randomized controlled trial, N=15,000, gamified intervention, pre/post/follow-up design

Analytic thinking (CRT) is the strongest individual-difference predictor of ability to discern real from fake news, outperforming political knowledge, media literacy, and partisan identity. Each additional CRT point associated with ~15% improvement in discernment.

Direction: positive Confidence: strong Method: Cross-sectional survey, N=2,739, headline discernment task, hierarchical regression

Propositions

Repeated misinformation becomes fluent, and fluency is misattributed to truth — this processing-based mechanism partially explains why retracted misinformation persists even when the retraction itself is remembered.

From: illusory_truth_effect To: continued_influence_effect Direction: positive

Prebunking (inoculation) is more effective than debunking because it builds resistance before exposure rather than trying to undo damage after encoding. Prevention outperforms cure for belief updating.

From: inoculation_effectiveness To: correction_effectiveness Direction: positive

Higher analytic thinking reduces susceptibility to the continued influence effect by promoting more careful evaluation of information sources and coherence checking during inference.

From: analytic_thinking To: continued_influence_effect Direction: negative

Playbooks

Quick Start
0 steps

Engines

meta_analysis logistic_regression ols_regression randomized_controlled_trial hierarchical_regression

Tags

topicmisinformation

Details

Domain: Misinformation & Belief Persistence

Study of how false beliefs form, persist after correction, and resist updating. Spans cognitive psychology (memory-based accounts), communication science (framing and source effects), and computational social science (spread dynamics). Core puzzle: why do people continue to rely on information they have been told is false?

Temporal scope: 1994-present | Population: Experimental participants (mostly WEIRD), social media users, survey panels

Key Findings

  • Corrections reduce but do not eliminate the continued influence effect. Even when participants recall and accept the correction, they continue to make inferences consistent with the retracted misinformation. Average residual influence: 20-40% of original misinformation effect persists. (positive, strong)
  • Meta-analysis of debunking studies (k=65, N>30,000) finds corrections are most effective when they provide an alternative causal explanation, not just a negation. Corrections with alternatives reduce CIE by 52%; simple negations reduce it by only 20%. (positive, strong)
  • The illusory truth effect occurs even for statements participants initially rated as false, and even when the source is explicitly labeled as low-credibility. Repetition increases perceived truth by d=0.35 on average. The effect is robust across labs, materials, and populations. (positive, strong)
  • The backfire effect does not replicate. Across 10,100 subjects and 52 contentious issues, corrections reduced misperceptions in every case. Not a single issue showed a statistically significant backfire. The original Nyhan & Reifler finding was likely a false positive. (null, strong)
  • Inoculation via the Bad News game reduced perceived reliability of misinformation by 21% compared to control, with effects persisting at 2-month follow-up. Technique-based inoculation (teaching manipulation strategies) generalizes better across topics than issue-based inoculation. (positive, strong)
  • Analytic thinking (CRT) is the strongest individual-difference predictor of ability to discern real from fake news, outperforming political knowledge, media literacy, and partisan identity. Each additional CRT point associated with ~15% improvement in discernment. (positive, strong)

Theoretical Propositions

  • [+] Repeated misinformation becomes fluent, and fluency is misattributed to truth — this processing-based mechanism partially explains why retracted misinformation persists even when the retraction itself is remembered.
  • [+] Prebunking (inoculation) is more effective than debunking because it builds resistance before exposure rather than trying to undo damage after encoding. Prevention outperforms cure for belief updating.
  • [−] Higher analytic thinking reduces susceptibility to the continued influence effect by promoting more careful evaluation of information sources and coherence checking during inference.

Installation

Install this PAX into your Praxis instance:

praxis_import_pax("misinformation-belief-persistence.pax.tar.gz", install=True)