Cognitive Biases, Systematically
A cognitive bias is a systematic pattern of deviation from rational judgment. The word 'systematic' is critical: biases are not random noise; they are predictable, directional, and replicable across people and situations. By the 2010s, researchers had catalogued over 180 named biases, prompting legitimate questions about whether the field had gone too far — naming every deviation without a coherent explanatory framework. This lesson takes a different approach: instead of listing biases, we organize them by mechanism. Understanding why a bias exists makes it far more useful than merely knowing its name.
Confirmation Bias: Seeking What Confirms
Confirmation bias is the tendency to search for, interpret, and remember information in a way that confirms existing beliefs. It operates at every stage of information processing: in what we look for (selective search), in how we evaluate what we find (selective interpretation), and in what we retain (selective memory). Peter Wason's 2-4-6 task is the cleanest laboratory demonstration. Participants are told that the sequence 2-4-6 conforms to a rule the experimenter has in mind. They can generate their own sequences and ask whether they conform to the rule, then guess the rule. Most people hypothesize 'even numbers increasing by 2' and then generate sequences like 4-6-8, 8-10-12, and 20-22-24 — all of which are confirmed. They almost never test sequences that would disconfirm their hypothesis (like 3-5-7 or 1-2-3). The actual rule is simply 'any three numbers in increasing order.' Participants test only confirming cases and therefore miss the rule. In the real world, confirmation bias shapes how we read news (we share articles that confirm our views and dismiss sources that challenge them), how we form opinions about people (we notice their behaviors that match our initial impression), how doctors diagnose (they anchor on the first hypothesis and underweight evidence pointing elsewhere), and how scientific communities resist paradigm shifts. It is one of the most robustly replicated findings in all of psychology.
Confirmation bias likely has evolutionary roots: in a stable environment, beliefs that have been tested and survived are more likely to be true than new, untested alternatives. It also reduces cognitive load — you do not have to re-evaluate everything from scratch every time you encounter new information. The bias becomes dangerous specifically in environments that are changing, where old beliefs may be outdated, or in complex domains where the initial hypothesis was formed on insufficient evidence.
Overconfidence bias is the tendency to overestimate the accuracy of one's own judgments, knowledge, and predictions. Three distinct phenomena fall under this heading. Overplacement is believing you are better than others: studies consistently find that 70-80% of drivers rate their skill as above average — a mathematical impossibility. Overprecision is believing your estimates are more accurate than they are: when people give 90% confidence intervals for facts (ranges they are 90% sure contain the true answer), the true answers fall within those ranges only about 50% of the time. Overestimation is thinking you will perform better than you actually do. Overconfidence is particularly pronounced in domains of low predictability — the economy, politics, complex social systems — where the feedback loops are too long or ambiguous to calibrate intuitions. Financial analysts, political forecasters, and medical experts all exhibit systematic overconfidence precisely where the environment does not provide rapid, clear feedback. Phillip Tetlock's landmark study of expert political forecasters found that across 82,361 predictions made by 284 experts over 20 years, experts were only marginally better than chance, and that those who expressed the most confidence were not the most accurate.
Framing Effects: The Same Fact, Different Choices
Framing effects occur when logically equivalent information presented in different ways produces different decisions. The classic demonstration is Kahneman and Tversky's 'Asian disease problem': Condition A: 'A disease will kill 600 people. Program A saves 200 lives.' Most people prefer Program A. Condition B: 'A disease will kill 600 people. Program B results in 400 deaths.' Most people reject Program B. Programs A and B are mathematically identical: in both cases, 200 people live and 400 die. The difference is framing: Condition A frames the outcome as a gain (lives saved); Condition B frames it as a loss (deaths). Prospect theory, developed by Kahneman and Tversky, explains this with loss aversion: losses loom psychologically larger than equivalent gains. The same outcome feels worse when framed as losing than as failing to gain. Framing effects pervade real decisions. Medical consent forms that describe a procedure as having a '90% survival rate' produce different acceptance rates than forms describing a '10% mortality rate.' 'Ground beef that is 25% fat' is judged less favorably than '75% lean.' Political rhetoric frames taxation as 'the government taking your money' or 'collective investment in public goods' — not because the factual content differs but because the frame activates different associations and emotional responses.
Hindsight bias is the tendency, after learning an outcome, to believe you would have predicted it. 'I knew it all along' — even when you did not. Baruch Fischhoff's original 1975 experiments showed that people who were told the outcome of historical events (e.g., the result of the British-Gurkha war) rated that outcome as having been much more predictable than did people who were not told the outcome — and showed no awareness that the information had changed their judgment. Hindsight bias has serious consequences in domains where post-event evaluation matters. Medical malpractice cases are distorted by it: juries evaluating a physician's decision after a bad outcome judge that decision as obviously negligent far more harshly than they would have judged the same decision made under the original uncertainty. Historical analysis is contaminated by it: analysts explain why major historical events (wars, economic crises) were 'inevitable,' failing to reconstruct the genuine uncertainty that decision-makers faced. And it is self-protective: by rewriting our past predictions toward the actual outcome, we preserve the illusion of knowing the world better than we do.
Match each cognitive bias to the core mechanism that drives it.
Terms
Definitions
Drag terms onto their definitions, or click a term then click a definition to match.
Structural Biases: Status Quo, Sunk Cost, and Attribution
Beyond the above, several biases arise from how we relate to commitments, change, and other people. Status quo bias is the preference for the current state over change, even when change would be beneficial. It is partly loss aversion (the loss of changing feels larger than the gain) and partly omission bias (doing nothing feels less blameworthy than acting and getting a bad outcome). It explains why people keep insurance policies, subscriptions, and relationships past the point of rational justification. Sunk cost fallacy is continuing to invest resources in a failing course of action because of prior investment. You have spent two hours waiting in a terrible restaurant and stay because you already waited — even though the two hours are gone regardless and staying makes the situation worse. Rationally, only future costs and benefits should determine current decisions. Sunk costs are irrelevant; that the mind treats them as relevant is the fallacy. Fundamental attribution error is the tendency to attribute others' behavior to their character or personality while attributing our own behavior to situational factors. When someone cuts us off in traffic, we conclude they are reckless; when we cut someone off, it is because we were late and distracted. This asymmetry in explanation preserves self-image while systematically misjudging others.
The research on cognitive biases is sometimes reduced to a list of catchy names — the IKEA effect, the decoy effect, the cheerleader effect. Knowing the name of a bias does not mean you understand its mechanism or will avoid it. And knowing the mechanism does not guarantee you will recognize it in your own thinking in the moment. Systematic improvement requires practice, institutional design, and external checks — topics covered in Lesson 8.
A company continues funding a failing product line because they have already invested $5 million in it. A rational analysis shows the product will never be profitable. This decision is best explained by:
A surgeon explains that a procedure has a 95% success rate. A different surgeon describes the same procedure as having a 5% failure rate. Research would predict: