Noticing what fits prior expectation and filtering out the rest — the perceptual stage of Confirmation Bias. The disconfirming item is not discounted in analysis; it is never registered as evidence at all.
In intelligence analysis
- Contradicting reporting is present in the file and absent from the product: nobody decided to exclude it, in a sense nobody saw it.
- It operates at collection and triage as well as at analysis — indicators are selected, feeds are tuned, and the filter is inherited from the working hypothesis rather than stated.
- It is why more data does not correct bias on its own: additional material is screened by the same expectations that produced the error, so the picture can become more confident as it becomes more selective.
- In monitoring and detection work it presents as tuning thresholds until the alerting matches expectations — the tool is calibrated to the hypothesis.
Countermeasure
- Search deliberately for the disconfirming item before drafting: require at least one piece of evidence that argues against the leading judgement, and cite it in the product.
- Have the alternate reading of the same evidence argued out loud (Structured Analytic Techniques).
Related
- Cognitive Bias — the taxonomy this bias sits within
- Confirmation Bias — the reasoning-level counterpart
- Vividness Bias — the striking item is perceived first and weighted hardest
- Structured Analytic Techniques — devil's advocacy exists to surface the unseen