{
  "id": "hasty-generalization",
  "title": "Hasty Generalization",
  "lang": "en",
  "source_lang": "en",
  "summary": "Hasty generalization is an inferential error in which a broad conclusion rests on evidence too limited or unrepresentative to justify it.",
  "category": "philosophy",
  "subcategory": "metaphysics-epistemology",
  "type": "concept",
  "aliases": [
    "Hasty generalisation",
    "Fallacy of hasty generalization",
    "Overhasty generalization"
  ],
  "generated_at": "2026-11-05",
  "model": "pro/gpt-6.1-sol",
  "url": "https://aiwiki.page/en/hasty-generalization/",
  "translations": {
    "zh-Hans": {
      "url": "https://aiwiki.page/zh/hasty-generalization/",
      "json": "https://aiwiki.page/zh/hasty-generalization.json"
    }
  },
  "sections": [
    {
      "heading": "",
      "text": "Hasty generalization is a fallacy in which a conclusion about a population, category, or recurring pattern is drawn from evidence insufficient to support its scope or certainty. It commonly involves treating a few observations, an unusual case, or a selectively gathered sample as representative of a much larger whole. The error concerns the relationship between evidence and conclusion: the observations may be accurate, and the conclusion may even happen to be true, without the observations adequately justifying it."
    },
    {
      "heading": "Structure of the inference",
      "text": "Hasty generalization is usually a defective form of inductive reasoning. A typical argument has this structure:\n\n- Some observed members of a group have a particular property.\n- Therefore, all, most, or a specified proportion of the group have that property.\n\nFor example:\n\nTwo appliances bought from a manufacturer failed shortly after purchase. Therefore, most of that manufacturer's appliances are unreliable.\n\nThe failures provide evidence about those appliances. Without additional information about the manufacturer's total output, product range, and failure rates, they do not establish the broader claim.\n\nUnlike deductive reasoning, induction does not generally guarantee its conclusion. Consequently, an inductive argument is not fallacious simply because its conclusion could be false. The relevant question is whether the evidence makes the conclusion sufficiently credible, given how broadly and confidently it is stated. “These failures suggest a possible reliability problem” is weaker than “most of the manufacturer's appliances are unreliable” and requires less support.\n\nThe defect is therefore not merely a lack of deductive validity. It is inadequate inductive support."
    },
    {
      "heading": "What makes a generalization hasty?",
      "text": "Insufficient sample size\n\nA small sample can give an unstable picture of a population because chance variation has a large influence on the result. Observing three satisfied customers, for instance, provides little basis for a precise estimate of satisfaction among thousands of customers.\n\nThere is no universal number of observations that separates sound generalization from hasty generalization. The required evidence depends on the population's variability, the sampling method, the precision sought, and the strength of the conclusion. A sample adequate for identifying a tentative pattern may be inadequate for estimating a rare event.\n\nUnrepresentative selection\n\nA large sample can still be misleading if the process used to select observations systematically favors some cases over others. This is a problem of sampling bias.\n\nFor example, a survey about library services conducted only among visitors to the library excludes people who do not use it. Its results may describe participating visitors well while failing to describe the wider community.\n\nIncreasing the number of observations collected by the same biased method does not necessarily correct the problem. Sample size and representativeness are distinct considerations.\n\nDependence among observations\n\nNumerous observations may contain less independent information than their number suggests. Several reports might repeat the same original account, or many defective items might come from a single production batch.\n\nTreating such observations as independent evidence can exaggerate the support for a general conclusion. The relevant issue is not just how many observations exist, but how they were generated and how closely they are related.\n\nExcessive scope or confidence\n\nA conclusion may extend beyond the place, period, conditions, or category actually studied. Evidence about one branch of a business does not automatically establish a claim about every branch; evidence gathered under controlled conditions does not automatically establish performance in every practical setting.\n\nHastiness can also lie in presenting a tentative inference as certain. Words such as “all,” “always,” and “never” impose particularly demanding evidential requirements, although probabilistic claims such as “usually” or “most” can also be inadequately supported."
    },
    {
      "heading": "Illustrative cases",
      "text": "Observation | Unsupported generalization | Evidential limitation\nThree students in one class dislike an examination format. | Students generally dislike that format. | The sample is small and restricted to one class.\nMost respondents to a voluntary online poll favor a proposal. | Most residents favor the proposal. | Participants select themselves into the sample.\nOne product performs well in a demonstration. | Every unit of that product will perform equally well. | A single demonstration does not establish consistency.\nSeveral news reports describe similar incidents. | Such incidents are typical or increasingly common. | Report selection and changes in coverage may not reflect population frequencies.\n\nThese examples identify gaps in support, not proof that the generalizations are false. Additional evidence could confirm, qualify, or overturn them."
    },
    {
      "heading": "Related concepts and distinctions",
      "text": "Anecdotal evidence. An anecdote describes a particular experience or event. It can establish that something occurred and can suggest a hypothesis worth investigating. Hasty generalization arises when such evidence is made to support a broader conclusion than it warrants. Anecdotes are therefore not inherently fallacious.\n\nStereotyping. Some stereotypes involve generalizing from limited encounters to an entire social group. However, the concepts are not identical. Stereotyping can also involve applying a group-level claim indiscriminately to individuals, whereas hasty generalization concerns inadequate support for the general claim itself.\n\nCherry-picking. Cherry-picking involves selecting favorable evidence while neglecting relevant contrary evidence. It can produce a hasty generalization, but the two describe different defects: selective presentation and insufficient support for generalization. Neither label alone establishes whether the selection was deliberate.\n\nCausal inference. Concluding from a few events that one factor causes another can combine hasty generalization with an error about causation. The post hoc fallacy specifically treats temporal succession as sufficient evidence of causation; it need not involve generalization from a small sample.\n\nThe problem of induction. The problem of induction concerns the general philosophical justification for reasoning beyond observed cases. Hasty generalization is a more local criticism: even within ordinary inductive practice, a particular conclusion exceeds the available evidence."
    },
    {
      "heading": "Statistical treatment",
      "text": "In statistics, the relationship between a sample and a population is addressed through sampling design, estimation, and explicit assumptions. A confidence interval can express sampling uncertainty under a specified statistical procedure, but it does not automatically account for biased selection, poor measurement, or an incorrectly defined target population.\n\nFor an idealized sample of \\(n\\) independent observations from the same population, where each observation records the presence or absence of a property, the sample proportion \\(\\hat p\\) has standard deviation\n\n\\[\\sqrt{\\frac{p(1-p)}{n}},\\]\n\nwhere \\(p\\) is the population proportion. This illustrates why larger independent samples generally reduce random sampling variation. It does not show that a large convenience sample is representative.\n\nStatistical rigor therefore requires more than accumulating cases. It also requires a defensible connection between the observations, the population of interest, and the particular claim being made."
    },
    {
      "heading": "Limits of the diagnosis",
      "text": "Small samples do not automatically make an argument fallacious. A single well-established counterexample can refute a universal claim: one nonwhite swan is enough to disprove “all swans are white.” That inference does not attempt to estimate the prevalence of nonwhite swans.\n\nLikewise, one observation may establish an existential claim—“at least one such object exists”—without supporting a generalization about how common it is. A small number of carefully controlled observations may also provide strong evidence for a narrowly specified conclusion when supported by relevant background knowledge.\n\nIdentifying hasty generalization therefore requires specifying the conclusion and explaining why its evidence is inadequate. Merely calling a sample “small,” or rejecting a conclusion because it was reached quickly, does not establish the fallacy."
    }
  ],
  "links": [
    "fallacy",
    "evidence",
    "inductive-reasoning",
    "deductive-reasoning",
    "logical-validity",
    "hypothesis",
    "causation",
    "post-hoc-fallacy",
    "problem-of-induction",
    "statistics",
    "confidence-interval"
  ],
  "sources": [],
  "license": "https://creativecommons.org/licenses/by/4.0/"
}