Inductive generalisation and sampling: QCE Philosophy and Reason
“Evaluate inductive generalisations by assessing sample size, representativeness and the dangers of hasty generalisation and biased sampling”
An inductive generalisation infers a feature of a whole population from a sample, so it is assessed for strength, not validity. A strong generalisation needs an adequate sample size and, more importantly, a representative sample, usually secured by random selection. Hasty generalisation draws a sweeping conclusion from too small or atypical a sample (often a single anecdote), while biased sampling uses a systematically unrepresentative sample through self-selection, convenience or survivorship bias. The 1936 Literary Digest poll shows that a huge but biased sample can be beaten by a smaller random one, and Hume's problem of induction shows even representative samples cannot guarantee the future.
Jump to a section
What this dot point is asking
QCAA wants you to evaluate the inductive move from a sample to a population: from "most people I surveyed prefer X" to "most people prefer X." You need the structure of an inductive generalisation, the two criteria that make it strong (adequate size and representativeness), and the fallacies that wreck it (hasty generalisation and biased sampling). This is essential for analysing scientific, statistical and everyday claims.
The answer
The structure of a generalisation
An inductive generalisation infers a claim about a whole population from observations of a sample of it:
- In the observed sample, proportion p of members have feature F.
- Therefore, in the whole population, roughly proportion p have feature F.
The conclusion goes beyond the evidence (we have not checked every member), so the argument is inductive and assessed for strength, not validity. Polling, quality control and most scientific measurement rest on this pattern.
Criterion one: sufficient sample size
A generalisation is stronger when the sample is large enough. Inferring the eating habits of a nation from three friends is weak; the small sample could easily be unrepresentative by chance. Larger samples reduce the role of luck. But size alone is not enough, which is why the second criterion matters more.
Criterion two: representativeness
A sample is representative when it resembles the population in the features relevant to the conclusion. A sample can be huge yet biased. The classic case is the 1936 Literary Digest poll, which surveyed millions of Americans drawn from telephone and car-registration lists and wrongly predicted the presidential result, because in the Depression those lists over-represented wealthier voters. A smaller but properly random sample beat it. Representativeness usually requires random selection so that every member of the population has an equal chance of being included.
Hasty generalisation
The hasty generalisation (or fallacy of insufficient sample) draws a broad conclusion from a sample that is too small or atypical. "My grandfather smoked and lived to ninety, so smoking is harmless" generalises from a single anomalous case. Anecdotes are the most common form: one vivid story is treated as evidence about everyone.
Biased sampling
Biased sampling uses a sample that is systematically unrepresentative. Two frequent forms:
- Self-selection: people who choose to respond differ from those who do not. Online reviews over-represent the very pleased and the very angry.
- Convenience sampling: surveying whoever is easiest to reach (your own friends, one suburb) rather than a random cross-section.
A famous related error is survivorship bias: studying only the cases that "survived" a process. In World War II, examining only returning aircraft for bullet holes ignored the planes that did not return, which is where the fatal damage actually was.
Evaluating a generalisation in a response
To assess a generalisation in QCAA style: (1) identify the sample and the target population; (2) ask whether the sample is large enough; (3) ask, more importantly, whether it is representative, looking for self-selection, convenience and survivorship bias; (4) judge the strength accordingly. Naming "hasty generalisation" earns little; you must explain why the sample fails to support the population claim.
Claim: "I asked twelve people in the university library on a Tuesday morning, and ten said they read for pleasure daily. So most Australians read for pleasure daily."
Step 1: Identify the sample and the target population
Sample: twelve people in a university library on a Tuesday morning. Target population: all Australians.
Step 2: Assess sample size
Twelve is very small, so chance alone could make the result unrepresentative. This already weakens the inference.
Step 3: Assess representativeness
This is the bigger problem. Library users on a weekday morning are not a cross-section of Australia: they are skewed toward students and committed readers (convenience sampling). The relevant feature (reading habits) is exactly where this group differs from the population.
Step 4: Name the fallacy and judge strength
The argument commits a hasty generalisation built on a biased convenience sample, so it is very weak. A strong version would need a larger, randomly selected national sample.
A large sample can still be badly biased, as the 1936 Literary Digest poll showed. The decisive question is not only how many were sampled but whether the sample resembles the population in relevant respects. A small random sample can be stronger than a huge self-selected one.
Where generalisation sits in philosophy
Inductive generalisation is the engine of empirical science and the target of Hume's problem of induction: even a perfectly representative past sample cannot logically guarantee the future. It also connects to the informal fallacies strand, since hasty generalisation is a fallacy of presumption. Evaluating it well is the difference between evidence-based reasoning and prejudice dressed up as data.
- Equating large with reliable
- A large self-selected sample is still biased; representativeness matters more than size.
- Ignoring how the sample was gathered
- Self-selection, convenience and survivorship bias can corrupt even careful-looking studies.
- Generalising from anecdotes
- A single vivid case is a hasty generalisation, not evidence about a population.
- Naming the fallacy without explaining it
- You must show why the sample fails to represent the population, not just label it.
Try this
Q1. State the two main criteria for a strong inductive generalisation. [2 marks]
- Cue. Sufficient sample size and representativeness of the sample.
Q2. Explain why a large sample can still produce a weak generalisation. [3 marks]
- Cue. If the sampling method is biased (self-selection, convenience), size does not fix the unrepresentativeness.
Q3. Identify the fallacy: "Two players from that school cheated, so the whole school is dishonest." [2 marks]
- Cue. Hasty generalisation from too small and atypical a sample.
Exam-style questions
Questions in the style of QCAA exam questions on this dot point, each with a worked answer. They are written by ExamExplained unless tagged "Past paper"; the year shows the paper a question is modelled on.
Practice6 marksEvaluate the following inductive generalisation: "Our website asked visitors to rate the new policy. Of the 5,000 who responded, 78 per cent approved. Therefore most citizens approve of the policy." Refer to sample size and representativeness.Show worked answer →
A 6 mark response identifies the sample and population, then weighs size against representativeness.
Sample and population. Sample: 5,000 self-selected website visitors who chose to respond. Target population: all citizens.
Sample size. 5,000 is large, which on its own would reduce chance error, so size is not the weakness here.
Representativeness (the decisive point). The sample is self-selected: only people who visit the site and choose to respond are counted, and the strongly opinionated respond most. It also excludes citizens who are offline or unaware of the poll. So the sample is systematically unrepresentative of all citizens (biased sampling), and a large biased sample (compare the 1936 Literary Digest poll) does not support a population claim.
Verdict. The generalisation is weak: representativeness, not size, is the problem.
Markers reward identifying self-selection bias, the size-versus-representativeness point, and a clear verdict.
Practice4 marksExplain survivorship bias and how it can undermine an inductive generalisation, using an example.Show worked answer →
A 4 mark response defines the bias and shows its effect on a generalisation.
Definition. Survivorship bias occurs when a sample includes only the cases that "survived" some selection process, omitting those that did not, so the sample misrepresents the whole population.
Effect on a generalisation. Conclusions drawn only from survivors generalise from an unrepresentative sample. Example: in World War II, examining bullet holes only on returning aircraft suggested reinforcing the hit areas, but the planes shot in the fatal areas never returned, so the true vulnerable spots were missing from the sample. Generalising from survivors gave the wrong conclusion.
Markers reward an accurate definition and an example showing how omitting non-survivors distorts the population inference.