Philosophy › Section E · Logic and reasoning › Topic E.4

Deductive, inductive, and abductive inference

Some reasoning aims at certainty, some at probability, and some at the best available explanation. Knowing which kind you are using tells you how to test it and how far to trust its conclusion.

Three kinds of inference

An inference is the step from premises to a conclusion. A deductive inference aims to guarantee its conclusion: if the premises are true, the conclusion must be. An inductive inference aims to make its conclusion probable, going beyond the information in its premises. An abductive inference reasons from evidence to the hypothesis that would best explain it. Deductive arguments are judged valid or invalid, as the topic "Arguments, validity, and soundness" explained. Inductive and abductive arguments are judged strong or weak, and strength comes in degrees.

Valid deductive forms

Some argument forms are valid whatever statements fill them. Modus ponens: if P then Q; P; therefore Q. Modus tollens: if P then Q; not Q; therefore not P. Disjunctive syllogism: P or Q; not P; therefore Q. Hypothetical syllogism: if P then Q; if Q then R; therefore if P then R. The conditions from the topic "Logical statements" show why the first two work. In "if P then Q", P is sufficient for Q, so P gives you Q. And Q is necessary for P, so without Q there can be no P.

Two invalid look-alikes

Two invalid forms resemble the valid ones. Affirming the consequent: if P then Q; Q; therefore P. "If the fire alarm is triggered, the guards are alerted. The guards were alerted. So the alarm was triggered." A phone call could have alerted them: the necessary condition holding does not guarantee the sufficient one. Denying the antecedent: if P then Q; not P; therefore not Q. "If a stone is a diamond, it scratches glass. This stone is not a diamond. So it does not scratch glass." Quartz is not a diamond, yet it scratches glass. In each case a single counterexample proves the form invalid.

Worked example: naming the form

1. "If the reactor overheats, the alarm sounds. The alarm did not sound. So the reactor did not overheat." The consequent is denied, so this is modus tollens: valid. 2. "The key is either in my bag or in the car. It is not in my bag. So it is in the car." Disjunctive syllogism: valid, though only as good as the claim that there are no other places. 3. "If Ana trains daily, she will finish the marathon. Ana finished. So she trained daily." Affirming the consequent: invalid, since she might have finished on talent or luck. 4. "If you are in Lyon, you are in France. If you are in France, you are in Europe. So if you are in Lyon, you are in Europe." Hypothetical syllogism: valid.

Inductive inference

Inductive conclusions go beyond their evidence. Common patterns include generalising from a sample ("every one of 200 tested bulbs lasted a year, so most bulbs of this type do"), applying a generalisation to a case ("most trains on this line run on time, so mine probably will"), reasoning from past to future and arguing by analogy. A strong inductive argument makes its conclusion probable if its premises are true; a cogent argument is strong and has true premises. Induction is defeasible: new evidence can weaken even a strong inductive argument, whereas adding premises never makes a valid argument invalid.

Strength and Hume's problem

How strong is strong enough? It depends on the stakes. A slightly better than even chance may justify a small bet, while a criminal conviction demands proof beyond reasonable doubt. Strength also depends on sample size, on whether the sample represents the whole, and on how relevant the evidence is. David Hume, in the work now known as An Enquiry Concerning Human Understanding (1748), raised a deeper worry, the problem of induction: every inductive inference assumes that the future will resemble the past, and that assumption cannot be proved deductively, while defending it inductively would be circular. Philosophers still debate replies, yet everyone relies on induction daily.

Abductive inference

Abduction, a name given by Charles Sanders Peirce, is also called inference to the best explanation, a phrase made popular by Gilbert Harman in 1965. It starts from evidence and infers the hypothesis that would best explain it. A mechanic who hears a rhythmic clicking that speeds up with the car infers a stone stuck in a tyre. Doctors diagnosing, detectives naming suspects and scientists proposing theories all reason this way. The direction differs from induction: in induction the premises are instances that support a conclusion, while in abduction the conclusion is meant to explain the premises. Like induction, abduction yields probability, not certainty.

Explanatory virtues

Rival explanations are compared by explanatory virtues. Explanatory power: it accounts for all the evidence, not just some. Simplicity: it assumes no more than it needs. This is Ockham's razor, after William of Ockham (about 1287 to 1347); the famous wording "entities must not be multiplied beyond necessity" is not found in his writings, though he stated similar principles. Conservatism: it fits well-established knowledge. Depth: it does not raise more questions than it answers. These are rules of thumb that can conflict, and the true explanation is sometimes the complicated one. A claim that overturns established knowledge needs correspondingly strong evidence: extraordinary claims require extraordinary evidence.

Worked example: choosing the best explanation

Your bicycle's back tyre is flat one morning. Evidence: a thorn is lodged in the tread, yesterday you rode down a hedge-lined lane, and the front tyre is fine. 1. Hypotheses: H1, the thorn punctured the tube; H2, someone let the air out; H3, the valve is faulty. 2. Explanatory power: only H1 accounts for the thorn and the lane. 3. Simplicity: H2 needs an unseen person and an unknown motive. 4. Conservatism: thorn punctures are common and well understood. 5. Test: pump up the tube and hold it under water. Bubbles at the thorn's position support H1; bubbles at the valve favour H3. 6. Conclusion: H1 is the best explanation for now, held provisionally until the test is done.

Common mistakes to avoid

First, treating affirming the consequent or denying the antecedent as valid because they resemble modus ponens and modus tollens. Second, calling an inductive argument "invalid" as a criticism; it never aimed at validity, so ask whether it is strong. Third, thinking every argument from general to particular is deductive: "most trains run on time, so mine will" is inductive. Fourth, treating the best available explanation as proven. Fifth, using Ockham's razor as if the simplest explanation must be true; it is a guide for choosing, not a guarantee.

Key takeaways

Deduction aims to guarantee its conclusion, induction to make it probable, abduction to explain the evidence. Modus ponens, modus tollens, disjunctive syllogism and hypothetical syllogism are valid; affirming the consequent and denying the antecedent are not. Inductive arguments are strong or weak, cogent when strong with true premises, and always open to new evidence; Hume argued that induction cannot be justified without circularity. Abduction compares rival hypotheses by explanatory power, simplicity, conservatism and depth, and its conclusions stay provisional.

Flashcards

Flashcard

Inference

The step from premises to a conclusion

Deductive, inductive or abductive.

Test your understanding

Question 1 of 5

"If the battery is dead, the car will not start. The car started. So the battery is not dead." Which form is this?

Question 2 of 5

"If a substance is table salt, it dissolves in water. This substance dissolves in water. So it is table salt." What is wrong with this argument?

Question 3 of 5

"The lights are off, the car is gone and post is piling up, so the neighbours are probably away." What kind of inference is this?

Question 4 of 5

What is a cogent argument?

Question 5 of 5

What does Ockham's razor recommend when comparing explanations?

← Previous topicNext topic →