Julia Galef · 2021
Why Some People See Things Clearly and Others Don't
Rough AI truth score
67/100
The practical core is well supported: actively open-minded thinking, calibrated confidence, and genuine engagement with counterevidence are associated with better reasoning. The evidence is less decisive about how easily these habits can be trained, whether self-reported scout identity causes accuracy, and how durable the gains are outside structured tasks.
Based on three central claims · medium confidence
Studies link actively open-minded thinking with seeking disconfirming evidence, reduced myside bias, and better calibration. Much of the evidence is correlational or task-bound, so causality and real-world effect sizes remain uncertain.
A large literature documents motivated reasoning and myside bias, although magnitude depends on identity strength, stakes, knowledge, incentives, and study design. Accuracy motives can counteract defensive processing but do not eliminate it.
Prompts, accuracy incentives, and debiasing exercises can improve performance in specific settings, but transfer and durability are uneven. Selection effects, social incentives, domain knowledge, and organizational culture may matter as much as an individual's stated mindset.
What this number means. It is an AI-generated first-pass judgment of three central factual or causal claims—not a rating, exhaustive fact-check, or human peer review. Claim credits are 100% for supported, 75% for mostly supported, 50% for mixed, and 25% for weak, then averaged and rounded. Lower confidence means the score should move more as better evidence arrives.
Method three-central-claims/0.1.0 · checked 2026-09-01 · 3/3 selected claims assessed · method and source audit