Based on data release 2026.08.18. This report separates observations from our corpus from claims established by external research.
Corrections usually help more than the backfire story suggests
Backfire is a memorable claim, but the reviewed evidence does not support treating it as the normal result of correction. The useful distinction is between disagreement, incomplete updating and a measured increase in the false belief.
Corrections usually improve factual accuracy. A true backfire effect, where a correction makes the targeted false belief stronger, appears to be uncommon and sensitive to context and measurement. It should not be treated as the default response to being corrected.
Retrospective judgment needs two different checks
Knowing the outcome can distort both perceived predictability and evaluation of the earlier decision. Keeping forecasts and decision rationales before the outcome gives reviewers something better than reconstructed memory.
Knowing an outcome can make the outcome look more predictable in retrospect. The effect has been studied for decades and across many settings, but it does not mean that every confident explanation after an event is biased.
Outcome bias occurs when knowledge of a result changes how people evaluate the quality of a decision even when the information available at the time of the decision is held constant. Outcomes can still be relevant for learning, so the error is not 'never look at results'; it is using luck or hindsight as if it had been available to the original decision-maker.
Repetition is not independent evidence
Repeated claims can feel more true even when repetition adds no new evidence. This matters especially when several pages or AI outputs are recycling the same source.
Repeated information is, on average, judged as more truthful than comparable new information. The size of the effect varies with the material and procedure, and repetition does not make every claim believable or erase all prior knowledge.
Uncertainty is a result, not a publishing failure
Mixed, contested and unreviewed states are useful outputs. They show where a confident internet summary would be stronger than the reviewed evidence.
What remains uncertain
This report summarizes the reviewed subset of the Cognitive Biases corpus. It is not a systematic review of every cognitive-bias study, and coverage is uneven across concepts.