A cognitive bias is a systematic — as opposed to random — error in judgement, driven by the mental shortcuts (heuristics) the brain uses to process information efficiently. In intelligence analysis the stakes are high: the same mechanisms that let an analyst make fast sense of a noisy stream also produce predictable, non-random mistakes.
Analysts do not passively receive information; they actively construct meaning through pre-existing mental models. Bias is therefore not occasional error but a built-in property of the perceptual process, largely unconscious and largely unaffected by how much information is in hand.
Why it matters
- Bias survives more data — additional information is filtered through the same model that produced the error.
- It is largely invisible to the analyst in the moment.
- It is the explicit rationale for Structured Analytic Techniques — every SAT is a structured counter to one or more biases (ACH against Confirmation Bias; Key Assumptions Check against unchallenged assumptions; Devil's Advocacy against Groupthink).
- Confidence calibration (Estimative Language) and source grading (Admiralty Rating) are also, in part, bias-mitigation instruments.
Taxonomy
Perception & mindset
- Confirmation Bias — seeking and weighting evidence that confirms an existing view; discounting disconfirming evidence.
- Mindset / belief perseverance — difficulty adopting a new frame once one is set.
- Mirror Imaging — assuming the other actor shares your values, rationality, and constraints.
- Selective perception — noticing what fits prior expectations and filtering the rest.
Evidence evaluation
- Vividness Bias — over-weighting concrete, emotionally striking material over abstract statistical information.
- Anchoring Bias — over-reliance on an initial figure or judgement; insufficient adjustment.
- Availability Heuristic — estimating likelihood by ease of recall rather than base rates.
- Representativeness heuristic — judging probability by stereotypical similarity, ignoring base rates.
- Absence-of-evidence fallacy — treating "no evidence found" as "evidence of absence".
Estimation & confidence
- Overconfidence Bias — systematically over-estimating the accuracy of one's own judgements.
- Hindsight Bias — retroactively judging an event as more predictable than it was.
- Base-rate neglect — ignoring prior probabilities in favour of vivid specifics.
- Insensitivity to sample size — drawing strong conclusions from small samples.
Social & group
- Groupthink — consensus pressure degrading rigorous appraisal (Janis 1972).
- Conformity / bandwagon — agreeing with the majority view regardless of one's own analysis.
- Fundamental attribution error — attributing others' actions to disposition, ignoring situational causes.
- Loss aversion (prospect theory) — losses loom larger than equivalent gains, skewing risk judgement.
Decision & commitment
- Sunk-cost fallacy / escalation of commitment — persisting with a failing line because of prior investment.
- Framing effect — choices reversed by how options are framed (gain vs loss).
- Status quo bias — preferring the current state.
- Recency bias — over-weighting the most recent reporting.
Countermeasures
- Structured Analytic Techniques — ACH, Key Assumptions Check, Devil's Advocacy, red-teaming.
- Explicit, calibrated Estimative Language.
- Admiralty Rating for source-level grading.
- Deliberate group process to counter Groupthink.
Related
- Structured Analytic Techniques — the primary mitigation toolkit
- Estimative Language — calibrated expression of confidence
- Admiralty Rating — source-level trust grading
- The Intelligence Cycle — bias operates chiefly in the evaluation/analysis stage