Generalization From Individual Cases: The Cognitive Bias That Drives Wrong Decisions

Generalization from individual cases — deriving rules, beliefs, and policies from specific personal experiences or single anecdotes — is a foundational error in human reasoning. It is also the most common form of evidence evaluation in everyday life, which is why its costs are so large.

"I know someone who smoked until 90 and was healthy" is the canonical example: a single case is treated as evidence that challenges the population-level association between smoking and mortality. This is not a logical argument — it is a demonstration of a cognitive bias.

The Statistical Issue

Base rate neglect: Humans systematically fail to weight base rates (population-level probabilities) appropriately when individual case information is available. A lawyer who looks confident and competent overrides the base rate that most people brought to trial for certain offenses are convicted. Your friend who lost weight on keto overrides the population-level evidence about long-term dietary adherence.

The statistical fact: individual cases are maximally uninformative about a question when compared to appropriately designed population studies. The reason is variance: any outcome that occurs in a distribution with high variance will have cases at both extremes — without that case being informative about the distribution's center or shape.

Availability heuristic: Cases that are vivid, emotionally salient, and personally proximate are remembered more easily and felt to be more representative than they statistically are. You remember the friend who got rich from cryptocurrency, not the many who lost money.

> 📌 Kahneman & Tversky (1973) in the foundational paper on availability and representativeness found that subjective probability is determined largely by the ease with which relevant examples come to mind — not by base rate information or sample size — leading to systematic distortions in risk estimation that cannot be corrected by simply telling people the correct statistics. [1]

Why This Bias Is Structurally Persistent

Individual cases carry information we've evolved to process: social information about specific other agents in our immediate group. Pre-statistical human environments had no meaningful "population-level data" — individual cases were the data. The brain's probabilistic reasoning evolved for this environment.

The statistical revolution of the last 400 years created a new type of information (population-level data from large samples with controlled variance) that our intuitive reasoning is poorly equipped to weight against vivid personal experience.

The Correctives

Ask what the population data shows before interpreting an individual case. The individual case may still be informative (as part of a pattern, as a hypothesis-generating signal), but it should be interpreted in light of, not instead of, population data.

Ask for the denominator. "I know someone who succeeded at X" — how many tried X? The numerator (successes you hear about) almost never arrives with the denominator (total attempts). This is survivorship bias compounded with availability bias.

Identify the mechanism. Individual cases that are anomalies (the 90-year-old smoker) do not disconfirm the mechanism — they represent the tail of a distribution the mechanism predicts most people will not occupy. Understanding why something causes harm (mechanism) is more informative than any individual case that seems to contradict the population evidence.

---

Key Terms

• Base rate neglect — the systematic failure to appropriately weight population-level frequency data when individual case information is available; the mechanism by which vivid personal examples override statistical evidence; documented by Kahneman & Tversky (1972, 1973)
• Availability heuristic — the cognitive shortcut estimating probability by how easily relevant examples come to mind; produces systematic overestimation of the frequency of vivid, emotionally salient, and recent events
• Anecdotal fallacy — the logical error of drawing general conclusions from a single or small number of specific cases without accounting for the variance of the distribution and the representativeness of the case
• Denominator blindness — the failure to consider the total population from which a visible outcome was drawn; produces systematic overestimation of success rates for visible outcomes (businesses that didn't fail, investments that paid off, treatments that worked)

---

Scientific Sources

• 1. Kahneman, D., & Tversky, A. (1973). On the psychology of prediction. Psychological Review, 80(4), 237–251. APA