Argument from Fallacy: Why a Bad Argument Does Not Make a Claim False

An argument has two separate properties: whether its premises support its conclusion, and whether its conclusion is true. Logic connects them in only one direction. If an argument is valid and its premises are true, the conclusion must be true. Nothing follows in the other direction: if an argument is bad, the conclusion may be true or false, and the argument has simply told you nothing about which. The argument from fallacy ignores this asymmetry and treats a failed argument as a successful argument for the opposite.

The mistake is easy to make because bad arguments are cheap and good ones are not. For almost any claim, true or false, somebody has argued for it badly. Gary Curtis's Fallacy Files makes the point with Fermat's last theorem: it is now accepted as true, it took more than three centuries to prove, and in the meantime many invalid arguments were offered for it. If a fallacious argument counted against its conclusion, the theorem would have been refuted long before it was proved. What is hard to find is a cogent argument for a claim, and that difficulty is the same whether the claim is true or false.

Psychology suggests why the two properties blur. Evans, Barston and Pollard showed in 1983 that people judge the validity of an argument partly by whether they believe its conclusion, accepting invalid arguments far more often when the conclusion seemed true. That is belief bias, and it runs from conclusion to argument. The argument from fallacy runs the other way, from argument to conclusion, but it rests on the same conflation: the sense that an argument and its claim are one thing, to be accepted or rejected together. Once the two are separated, the inference loses its grip.

There is a limit to the point, and Curtis states it. If every argument that has been offered for a claim fails, and the burden of proof lies with the claim's proponents, it is reasonable to reject the claim provisionally; an improbable proposition with no adequate support should not be believed. The fallacy is in the jump from one bad argument to the conclusion that no good one can exist, or from a weakness in the argument to a verdict on the claim. The tell is speed: the argument from fallacy settles a question in the time it takes to name a fallacy, and real questions rarely settle that fast.

The argument from fallacy is a product of logic rather than of the laboratory. The practice of naming the ways an argument can mislead goes back to Aristotle's Sophistical Refutations, which separated real refutations from apparent ones and set out thirteen kinds of apparent refutation, some depending on language and some not. Aristotle did not name the fallacy fallacy, but his project supplied the material for it: once there are cataloged fallacies to point at, there is a new way to go wrong, which is to treat the pointing as a refutation. The modern name, with its Latin alias argumentum ad logicam, belongs to the informal logic tradition, and the clearest statement of its form is in Gary Curtis's Fallacy Files: argument A for conclusion C is fallacious, therefore C is false.

The psychological evidence most closely related to it is on belief bias, the mirror-image confusion of argument with conclusion. Jonathan Evans, Julie Barston and Paul Pollard published "On the conflict between logic and belief in syllogistic reasoning" in Memory and Cognition in 1983. Across three experiments they crossed whether a syllogism was valid with whether its conclusion was believable, and asked students whether the conclusion followed. In the third experiment, 32 first-year psychology students each judged eight arguments under instructions stressing that only conclusions following by logical necessity should be accepted. Acceptance ran 91 percent for valid arguments with believable conclusions, 53 percent for valid arguments with unbelievable conclusions, 67 percent for invalid arguments with believable conclusions and 6 percent for invalid arguments with unbelievable conclusions. Students were judging the argument by whether they already accepted its conclusion.

The argument from fallacy is the same confusion in the opposite direction: judging the conclusion by the argument. The two errors share a cause, the treatment of an argument and its claim as a single unit, and the 1983 result shows how strong that treatment is even when people are told explicitly to judge logic alone. No experiment has measured the fallacy fallacy on its own terms, which is why the demo above walks through its logic rather than reproducing a study.

The pattern shows up wherever an argument is easier to examine than the claim it supports. The rows are illustrations, not findings from the cited studies.

The counter is a habit of treating an argument and its conclusion as two questions.

• Name what the fallacy removed. A fallacy takes one argument off the table. Say so explicitly, then ask what else bears on the claim, since the claim is still there.
• Ask whether a better argument exists. For most true claims a sound argument can be found; for most false ones it cannot. Spend a minute looking before concluding anything from the bad one.
• Separate the person from the proposition. A claim argued by someone careless, hostile or self-interested is not thereby false. Rewrite it in neutral words and judge the rewrite.
• Notice the pleasure of the catch. Spotting a fallacy feels like winning, and the feeling invites a verdict on the claim. Treat the feeling as a cue to slow down, not as a conclusion.
• Apply the burden of proof honestly. If every argument for a claim has failed and the claim is improbable on its face, provisional rejection is reasonable. If one argument has failed and the claim is otherwise plausible, the question is open.

The argument from fallacy is about the step from a bad argument to a false conclusion. Its neighbors are other ways of letting something beside the evidence decide the question.

• It is not belief bias, which runs the other way: judging whether an argument is valid by whether its conclusion is believable. The two share a cause, the treatment of argument and conclusion as one thing, but belief bias misjudges the logic and the argument from fallacy misjudges the claim.
• It is not the bandwagon effect, which is the pull toward believing what many others believe. The appeal to popularity in the demo is the bandwagon in argument form; the argument from fallacy is what happens when that appeal is rejected and the rejection is mistaken for a disproof.
• It is not confirmation bias, the preference for evidence that supports what one already thinks. Confirmation bias often decides which arguments get scrutinized for fallacies in the first place, but it concerns the search for evidence, not the inference from a failed argument.