Cognitive biases are automatic thinking mechanisms that divert our judgment and distort our decisions. Daniel Kahneman showed that our everyday choices are often influenced by a series of cognitive biases “of which we are unaware.” These mental distortions (often unconscious) push us to favor certain information, simplify problems, or rely on impressions rather than factual data.
Taxonomy of cognitive biases
Figure: Cognitive Bias Codex – a visual schema synthesizing more than 180 biases classified into four major categories. As the figure above shows, cognitive biases can be grouped according to the context in which they arise. We can distinguish, for example:
- biases related to information overload (the brain cannot process everything, so it uses fast heuristics),
- biases related to lack of meaning or context (a tendency to create patterns, stereotypes, or generalities from limited data),
- biases related to urgency to act (the need to decide quickly, which favors overconfidence and shortcuts), and
- memory biases (distortions of memory when recalling information).
For instance, memory biases are “distortions of judgment […] due primarily to memory effects”: the brain retains some details rather than others, forgets or modifies memories, and sometimes draws false generalizations. In each category, biases function as mental shortcuts—useful in the short term—that often backfire.
1) Information overload
When faced with an influx of data, the brain chooses simplicity (e.g., representativeness and availability heuristics).
- Definition: mental shortcuts used to process information quickly.
- Cause: limited capacity of System 1 (fast, intuitive thinking).
- Example: assuming a candidate who frequently performs well will always “score high” (overinterpreting a small sample, i.e., the “law of small numbers”).
- Mitigation: slow down the fast system (read carefully, verify statistics, consult experts).
2) Lack of meaning
When information is fragmented, we try to interpret it by creating stories or filling in gaps.
- Definition: tendency to perceive patterns or causes that don’t exist (e.g., illusion of correlation).
- Cause: System 1’s need for narrative coherence.
- Example: believing there is “necessarily a plot” behind isolated events (confabulation, stereotyping, appeal to authority).
- Mitigation: confront your hypotheses with other sources, test the logic behind each “story,” and ask the opposite question (“What if it isn’t true?”).
3) Urgency to act
Under time pressure or a desire for efficiency, we accept cognitive shortcuts.
- Definition: biases that emerge when we must decide quickly (e.g., overconfidence, Dunning–Kruger, status quo bias).
- Cause: desire to avoid uncertainty and remain in control.
- Example: unrealistic planning due to optimism (planning fallacy), or continuing a losing project due to commitment (sunk cost).
- Mitigation: introduce formal decision checkpoints (“coffee break” critical pause) and appoint a devil’s advocate for each decision.
4) Memory and reconstruction of the past
When we recall events or learn from experience, memories are reshaped.
- Definition: all memory biases (selective memory, recency effect, false memories).
- Cause: active reconstruction of the past by System 1.
- Example: favoring the most recent information (recency effect)—remembering the last candidates seen more vividly—or overestimating past successes based on subjective impression.
- Mitigation: keep notes (“logbook”), use objective tracking tools (budgets, quantified schedules), and regularly review past data.
Major cognitive biases (20+)
Confirmation bias: tendency to retain and interpret information in a way that confirms pre-existing beliefs.
- Cause: desire for internal consistency.
- Example: a recruiter convinced an “unusual” profile is good will seek only positive signals and ignore contradictions.
- Mitigation: actively search for counterexamples and request an outside view (devil’s advocate method).
Availability bias: estimating the likelihood of an event based on how easily examples come to mind.
- Cause: System 1’s reliance on salient information.
- Example: overestimating a highly publicized risk (e.g., plane crashes) because you saw many images, while underestimating more common risks (e.g., car accidents).
- Mitigation: consult real statistics; step back from what dominated recent information.
Anchoring bias (primacy effect): giving excessive weight to the first piece of information received.
- Cause: insufficient adjustment by System 2 after an initial anchor.
- Example: in negotiation, the first price proposed strongly shapes counteroffers.
- Mitigation: deliberately change the reference anchor, reframe the question, calculate independently.
Recency bias: favoring the most recent information.
- Cause: fresher memory trace in System 1.
- Example: preferring the last interviewee because their profile is still “fresh,” even if earlier candidates were stronger.
- Mitigation: evaluate candidates later with a clear head (random order or after a break); record impressions immediately for later comparison.
Halo effect: transferring a global positive (or negative) impression from one attribute to others.
- Cause: cognitive simplification that merges qualities.
- Example: judging a candidate competent because they are likable or have a prestigious résumé.
- Mitigation: evaluate each criterion independently (competency grid), use multiple evaluators.
Projection bias (false consensus): believing most people share our opinion.
- Cause: cognitive egocentrism.
- Example: assuming a technical choice is “obvious” and will be accepted by everyone.
- Mitigation: compare with objective data or collect diverse viewpoints.
Stereotype and association bias: generalizing based on groups or past experiences.
- Cause: category-based simplification (System 1).
- Example: evaluating an application negatively because it resembles a profile previously judged inadequate.
- Mitigation: check concrete evidence beyond the initial impression; promote diversity of perspectives (bias-awareness training).
Framing effect: the wording or context of a question influences the decision.
- Cause: System 1 sensitivity to wording and angles.
- Example: leading someone with a biased question (“You know Photoshop, so you’re creative, right?”).
- Mitigation: ask neutral questions; present the same option in multiple frames (gain vs. loss); rephrase to see whether the answer changes.
Contrast effect: comparison with a nearby reference distorts judgment.
- Cause: relative context dominates evaluation.
- Example: after several weak candidates, a recruiter overestimates the next candidate by contrast.
- Mitigation: assess each option independently; insert “standard” reference options.
Representativeness bias: judging probability by similarity to a prototype rather than statistical likelihood.
- Cause: categorization heuristic.
- Example: believing a broker is “necessarily” high-risk because they don’t fit the typical profile, ignoring base rates.
- Mitigation: check base rates; verify historical frequency of similar cases.
Availability heuristic: treating an event as more likely if it is recent or memorable (see “availability” above).
Barnum effect (Forer effect): accepting vague, flattering descriptions as highly accurate for oneself.
- Example: finding a generic personality test “uncannily personal.”
- Mitigation: recognize generalities; don’t rely on subjective descriptions alone.
Self-serving bias: attributing successes to oneself and failures to external factors.
- Example: taking credit for a good outcome; blaming circumstances for a bad one.
- Mitigation: conduct neutral post-mortems; analyze causes objectively.
Overconfidence bias: overestimating the accuracy of one’s predictions and skills.
- Example: forecasting an excessive profit margin due to blind confidence.
- Mitigation: apply safety margins; consult a mentor or second opinion.
Dunning–Kruger effect: less competent people overestimate their level; more competent people can be overly modest.
- Example: a novice believes they manage budgets well, while an expert highlights weaknesses.
- Mitigation: expert validation of estimates; encourage intellectual humility and calibration.
Status quo bias: preferring to keep the current state.
- Example: renewing an expensive contract out of fear of change even when a better option exists.
- Mitigation: explicitly analyze opportunity costs; set a scheduled “review date” to revisit the choice.
Planning fallacy: systematically underestimating the time or cost of a project (“optimistic plan”).
- Example: committing to an underestimated delivery date (often 30–50% too short).
- Mitigation: base plans on historical data; add buffers; break work into smaller tasks.
Escalation of commitment / sunk cost fallacy: continuing to invest in a losing project to avoid admitting past losses.
- Example: continuing product development despite catastrophic test results “so we don’t waste the money already spent.”
- Mitigation: use the “past-independent decision” rule (“What would we do if these costs didn’t exist?”); focus on future benefits.
Actor–observer bias: attributing our own behavior to situational causes but others’ behavior to personality traits.
- Example: explaining our lateness by traffic, while criticizing a colleague’s lateness as irresponsibility.
- Mitigation: consider others’ external constraints; apply the same interpretive standard to self and others.
Mere exposure effect (familiarity): liking something more simply because you’ve seen it before.
- Example: a recruiter favors a candidate met previously at an event because they feel familiar.
- Mitigation: anonymize non-relevant cues; compare neutrally (CV without photo/history).
In-group bias: unconsciously favoring members of one’s own group.
- Mitigation: form mixed teams for decisions; verify judgments aren’t driven by group lines (gender, age, culture).
Comparative table of biases
| Error (bias) | Category | Triggers (situation) | Impact | Likelihood | Key mitigation |
|---|---|---|---|---|---|
| Confirmation | Interpretive | Strong initial position, partial info | Locked decision, tunnel vision | Very high (nearly universal) | Actively search for a counterexample |
| Availability | Cognitive load | Recent/memorable representations | Over/underestimating risks | High (media, emotion) | Check real statistics |
| Anchoring (primacy) | Cognitive load | First number/value/info presented | Distorted estimate (near anchor) | Frequent (evaluation, negotiation) | Reframe with other reference points |
| Recency | Memory | Last info received | Earlier info neglected; unearned preference | Frequent (in-person, interviews) | Revisit older info impartially |
| Halo effect | Association | Strong overall impression (likability, prestige) | Incorrect generalization | Medium (interviews, reviews) | Evaluate each criterion independently |
| Framing | Wording | Gain vs. loss framing; context | Choice influenced by presentation | High (marketing, consultations) | Reframe the same option multiple ways |
| Contrast effect | Comparison | Sequentially very different options | Unfair comparison; biased choice | Medium (hiring, purchasing) | Insert standard reference options |
| Overconfidence (Dunning–Kr.) | Competence | Self-estimation of ability | Overestimation; overly certain judgment | Frequent (self-evaluation) | Regular feedback; expert calibration |
| Stereotypes / self-justification | Social perception | Cultural/social priors | Unconscious discrimination; judgment errors | Variable | Awareness training; diversity in selection process |
| Endowment effect | Motivation | Sense of ownership | Overvaluing owned assets | Fairly frequent | Recall market value; adopt an external perspective |
| Etc. | … | … | … | … | … |
Notes: “Category” refers to major classes (cognitive load, interpretation/meaning, social context, memory, etc.). “Likelihood” is indicative (synthesizing studies and observations): some biases (confirmation, availability) are close to universal, others (e.g., endowment) are more context-specific.
Quick checklist (individuals and teams)
For any decision-maker (you)
- Reflective pause: before deciding, step back (wait 10 minutes; switch tasks briefly).
- Challenge your intuitions: write your starting assumptions and imagine contrary data or opinions.
- Verify statistics: when possible, compare intuitions with facts (numbers, studies, benchmarks).
- Set “alarm signals”: flag sensitive decision points (“Sprint 1: doubt about the estimate”).
- Diversify information: consult multiple sources (colleagues, experts, users) to avoid a single-perspective bias.
For teams
- Create cognitive diversity: include varied backgrounds (gender, culture, discipline) to reduce in-group bias.
- Assign a devil’s advocate: each major proposal must be openly challenged by someone designated to question it.
- Discussion norms: encourage doubts and constructive critique; ban “because that’s how it is” decisions.
- Structured debate tools: use methods such as guided brainstorming, De Bono’s Six Thinking Hats, or anonymous reflection groups.
- Training and awareness: periodically train teams on biases so they are easier to spot (checklists, quizzes, fictional cases).
5-step decision checklist
- Clarify the goal and criteria: define the problem precisely, key stakes, and success indicators.
- Collect information and alternatives: gather reliable data and list all possible options (even the “crazy” ones).
- Search for potential biases: at each step ask: “Which bias could I be committing?” (anchoring, confirmation, availability, etc.). Record them in a decision journal.
- Evaluate each option objectively: weigh factual pros/cons. For each choice, imagine the opposite scenario (worst case), seek an independent opinion or predictive model.
- Document and review the decision: record reasons for the final choice and schedule a post-decision review (compare actual results to initial forecasts and adjust processes accordingly).
Practical case (timeline)
Context: a tech SME considers launching a new product. After an optimistic first meeting, the team follows a plan with excessive confidence.
- Day +0: Project kickoff — the PM sets an unrealistic launch date (planning fallacy) without consulting technical teams.
- Day +30 (Week 4): First demo — the team realizes the deadlines are far too short. They hesitate to adjust the plan (sunk cost / escalation of commitment).
- Day +45: Internal feedback — a senior engineer expresses doubt; the team resists (collective self-serving bias).
- Day +60 (Week 9): Progress meeting — a manager proposes applying the new decision checklist. The team revisits objectives, openly discusses fears (status quo bias), and decides to delay the launch.
- Day +75: Course correction — by adding time to critical steps (mitigating the planning fallacy) and outsourcing part of the work (seeking external input, countering confirmation bias), the project stays on track. Final launch occurs at Day +120 under the revised plan, with improved success rate.
This case illustrates why biases (overconfidence, sunk cost, conformity) must be identified and corrected through a formalized process (critical review, dashboard) rather than letting errors accumulate.
Decision process with bias checkpoints
Show code
This diagram illustrates a decision cycle integrating checkpoints (“bias review”) to detect and neutralize cognitive biases before finalizing a decision. At any time, if a bias is suspected, the process returns to analysis (Step C).
