Abduction
from 1878 Named by Peirce; the practice is as old as diagnosis
Reasoning backwards from something observed to whatever would best explain it, which is how diagnosis, detection and most of science actually proceed.
- Active
- from 1878
- Where
- Cambridge, Massachusetts
- Status
- In use
- Tradition
- Forms of inference
What it is
Abduction starts from a surprising fact and asks what would have to be true for it not to be surprising. Among the candidate explanations, it selects the best — and the whole difficulty of the method is packed into that word.
It is the only one of the three basic inference types that generates new hypotheses rather than testing or extending existing ones. Deduction unpacks, induction extends, abduction invents.
How it works
Note the surprising fact. Generate candidate explanations — as many as can be had, because the method can only choose among what it is given. Compare them for how completely they explain the fact, how much they require you to assume, and how well they fit what else is known. Adopt the best, provisionally, and go looking for the evidence that would tell it from its rivals.
The form of the inference
- The surprising fact F is observed
- If H were true, F would follow as a matter of course
- No rival explanation accounts for F as economically or as completely
- Therefore there is reason to suspect that H is true
The argument
Proposition The car will not start because the battery is flat.
- Turning the key produces a rapid clicking and nothing else, which is surprising in a car that ran yesterday.
- A flat battery would produce exactly that sound, as the starter solenoid engages and releases without power to turn the motor.
- A failed starter motor would more usually produce silence or a single heavy click; a fuel problem would let the engine turn over without firing, which is not what is happening.
- The interior light was left on overnight, which supplies a cause for a flat battery and none for the alternatives.
Therefore The battery is probably flat, and the next step is to test it rather than to trust the reasoning.
The last line matters. Abduction produces the hypothesis worth testing, not the finding. Anyone who stops here and orders a battery has mistaken the method for a conclusion.
What it is good for
Diagnosis of every kind — medical, mechanical, forensic — and the hypothesis-forming end of science. It is what you use when the evidence is in and the cause is not.
Where it fails
It can only choose among the explanations you happened to think of. If the true explanation never occurred to you, abduction will hand you the best of a bad set with the same confidence it would hand you the right answer, and nothing in the procedure will warn you.
It also depends on an unexamined notion of what makes one explanation better than another. Simpler, more familiar, more satisfying and more probable are four different criteria, they can point in different directions, and the method does not say which to follow.
Common misconceptions
That it is what Sherlock Holmes calls deduction. It is, and he is wrong about the name — nothing he does carries deductive necessity, which is why he is occasionally mistaken and why a real detective needs corroboration.
Also that the best explanation is thereby probably true. That holds only if the true explanation was among those considered, which is exactly what cannot be checked from inside the method.
Where it came from
Named and analysed by Charles Sanders Peirce in the 1870s, who regarded it as the only inference capable of producing a new idea and spent decades trying to work out what licenses it.
What became of it
Adopted widely under the plainer name inference to the best explanation, and now central to philosophy of science, epistemology and artificial intelligence.
Lasting influence
It underwrites scientific realism, most accounts of theory choice, and the everyday reasoning of every clinician and mechanic alive.
Why it still matters
The most-used method on this list outside formal contexts, and the least formally understood.
Put another way
Reading footprints. You never see the walker, and you can be very confident and quite wrong.
Texts
Charles Sanders Peirce, Collected Papers, volume 5
The abduction material is scattered and Peirce changed his mind more than once, which makes a good modern selection more useful than the volume itself.
Where to start
Peter Lipton's Inference to the Best Explanation is the standard modern treatment and is written to be read.
The surprising bit
Peirce thought abduction was not merely useful but the only inference that can enlarge knowledge at all, since deduction adds nothing and induction only extends what abduction first proposed.
In their own words
Abduction is the only logical operation which introduces any new idea
Who held it
Peter Lipton, Gilbert Harman
Connections
- Relies on
- Induction
- Influenced
- Probabilistic inference
- Parallel to
- Ockham’s razor
The verdict
Indispensable, unavoidable, and only as good as the imagination of the person using it.