Inference to the best explanation and abduction: QCE Philosophy and Reason
“Explain inference to the best explanation (abduction) and evaluate hypotheses using criteria such as simplicity, explanatory scope and coherence”
Inference to the best explanation, or abduction, infers that the hypothesis which best explains a surprising observation is probably true; it reasons backwards from evidence to its likely cause, unlike deduction or enumerative induction. An explanation is judged best by criteria such as explanatory scope, depth, simplicity (Ockham's razor), coherence with existing knowledge, testability and conservatism, with trade-offs requiring judgement. Abduction drives much of science, from Darwin's case for natural selection to the postulation of atoms and dark matter. It is fallible: rival hypotheses may be underdetermined by the evidence, and the best available explanation can still be the best of a bad lot.
Jump to a section
What this dot point is asking
QCAA wants you to understand a third mode of reasoning alongside deduction and induction: abduction, or inference to the best explanation (IBE). You infer that the hypothesis which best explains the evidence is probably true. You need the structure, the criteria used to judge which explanation is best, and the limits of the inference. IBE is how detectives, doctors and scientists actually reason from evidence to cause.
The answer
The structure of abduction
An inference to the best explanation has the form:
- Some surprising fact E is observed.
- Hypothesis H, if true, would explain E.
- No rival hypothesis explains E as well as H.
- Therefore H is probably true.
This differs from deduction (which guarantees the conclusion) and from simple enumerative induction (which generalises from instances). IBE reasons backwards from an observation to its best cause or explanation. Charles Sanders Peirce named this pattern abduction.
You return home to find the floor wet, the window open and a fallen vase. Decide which explanation is best.
Step 1: State the surprising evidence to be explained
E: three observations together (wet floor, open window, fallen vase). A good explanation should account for all of E, not just one part.
Step 2: List the rival hypotheses
H1: a burglar entered. H2: a freak gust blew the window open and knocked the vase, and the floor was already wet. H3: the cat came through the open window, knocked the vase (spilling its water onto the floor).
Step 3: Apply the criteria
Scope: H3 explains all three observations from one cause; H1 leaves the spilled water unexplained and predicts missing valuables (not observed); H2 needs two unrelated events. Simplicity: H3 posits one ordinary cause, satisfying Ockham's razor. Coherence: H3 fits background knowledge that the cat roams.
Step 4: Infer the best, holding it provisionally
H3 wins on scope, simplicity and coherence, so infer it as probably true. The inference is defeasible: discovering the cat was at the vet would defeat it, illustrating that the best explanation is not proven.
Criteria for the best explanation
Calling an explanation "best" is not arbitrary; philosophers list criteria a good explanation tends to satisfy:
- Explanatory scope: it accounts for more of the evidence than rivals.
- Explanatory depth / power: it explains the evidence in detail, not just vaguely.
- Simplicity (Ockham's razor): it posits fewer entities or assumptions. William of Ockham's principle says do not multiply entities beyond necessity.
- Coherence: it fits with what we already know rather than requiring us to abandon well-established beliefs.
- Testability: it makes further predictions that could be checked.
- Conservatism: it disturbs our existing web of belief as little as possible while still doing the job.
An explanation that scores well across these criteria is preferable; trade-offs between them (a simpler theory with less scope, say) are where the hard judgement lies.
Abduction in science and philosophy
Much scientific reasoning is abductive. Darwin argued that natural selection was the best explanation of a huge range of biological facts (the fossil record, vestigial organs, biogeography), not that it was deduced from them. The case for unobservable entities such as atoms or dark matter is abductive: they are posited because they best explain observed phenomena. IBE is also used in philosophy itself, for example in arguing that scientific realism best explains the success of science.
Limits of abduction
IBE is inductive and so fallible. Three cautions:
- Underdetermination: more than one hypothesis may fit the evidence equally well, leaving no clear "best."
- The best of a bad lot: the best available explanation may still be poor if we have not thought of the true one; "best so far" is not "true."
- Bias in the criteria: judgements of simplicity and coherence can be subjective and theory-laden.
Despite these limits, abduction is indispensable; we could not reason from evidence to causes without it.
Deduction moves from premises to a guaranteed conclusion and induction generalises from instances, but inference to the best explanation works backwards: it starts from a surprising observation and infers the hypothesis that would best account for it. The inference is only as good as the field of rival explanations actually considered.
- Treating the best explanation as proven
- IBE is inductive; the best available explanation can still be false.
- Ignoring rivals
- You cannot call an explanation best without comparing it to alternatives; otherwise you have only one explanation, not the best one.
- Using only simplicity
- Simplicity is one criterion among several; an overly simple theory may sacrifice explanatory scope.
- The best of a bad lot
- If the true explanation was never on the table, the "best" one chosen may be wrong.
Try this
Q1. State the structure of an inference to the best explanation. [3 marks]
- Cue. Surprising fact E; H would explain E; no rival explains E as well; so H is probably true.
Q2. Name and explain two criteria for judging which explanation is best. [4 marks]
- Cue. For example explanatory scope (accounts for more evidence) and simplicity (fewer assumptions, Ockham's razor).
Q3. Explain the "best of a bad lot" objection. [3 marks]
- Cue. The best explanation considered may still be false if the true one was never among the alternatives.
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 marksExplain inference to the best explanation, and use at least two criteria to evaluate which of two competing explanations of a surprising observation is better.Show worked answer →
A 6 mark response gives the structure of IBE and applies criteria to rank rivals.
Structure. A surprising fact E is observed; hypothesis H would explain E; no rival explains E as well; therefore H is probably true. The inference runs backwards from evidence to its best cause (abduction).
Worked comparison. Observation: a patient has a fever, rash and recent travel. H1: a common viral infection. H2: a rare tropical disease.
Apply criteria. Explanatory scope: if H1 leaves the travel history and specific rash unexplained while H2 accounts for all three, H2 has wider scope. Simplicity (Ockham's razor): H1 posits a more common cause, so it is simpler. Coherence and testability would then be weighed (a confirmatory blood test). The "best" explanation is the one that maximises scope and coherence at the least cost in assumptions.
Markers reward the IBE structure and the explicit use of at least two named criteria to rank the explanations.
Practice4 marksExplain the "best of a bad lot" objection to inference to the best explanation and one implication it has for how we should use abduction.Show worked answer →
A 4 mark response states the objection and draws an implication.
Objection. IBE infers the best of the explanations we have actually considered, but if the true explanation was never among them, the "best" one selected may still be false. Calling it best ranks it only within a possibly poor field.
Implication. We should treat the conclusion of an abductive inference as provisional and actively widen the field of hypotheses considered, rather than treating "best so far" as "true". It also warns against premature confidence when only one or two explanations have been canvassed.
Markers reward an accurate statement of the objection and a sensible methodological implication.