Shifting the burden of proof

The yard-sale chair

Seller: This chair was made by a famous furniture maker. That’s why it’s $400.

Buyer: How do you know? There’s no mark on it.

Seller: Can you show me it wasn’t?

The clear-cut case. The seller made the claim, and the claim is what sets the price, so the seller is the one who needs a reason: a maker’s mark, a receipt, a record of where it came from. The buyer can’t disprove it standing in a driveway, but that tells her nothing about who made the chair. It only shows that nobody present knows.

“Unless someone can show it won’t work”

At a team meeting, a manager proposes moving every project onto a new task-tracking tool. “Unless anyone can show me it won’t work for us, we’re switching on Monday.” Nobody has used the tool, so nobody can.

This one sounds fair, since anyone is free to object. But the manager hasn’t given a reason the tool will work for the team; he has set the default so that silence counts as agreement, and asked the others to prove a negative about something none of them has tried. Who should carry the burden in a decision like this can itself be argued over (see Burden of proof), but it has to be argued, not simply announced by the person who wants the change.

The lucky lure

Sam: This lure is lucky. We always catch more with it.

Ana: We caught nothing with it yesterday.

Sam: Yesterday would have been even worse without it. Prove it wouldn’t.

The less obvious form: a demand for disproof that no evidence could meet. Sam has shaped the claim so that a good day confirms it and a bad day is explained away, which makes it unfalsifiable. Asking Ana to disprove it asks for something impossible, and her failure then looks like support. An unmeetable demand for disproof is the burden shift at its most effective.

• “Prove me wrong.” An explicit demand that the listener disprove the claim, offered in place of support.
• Setting the default. Framing a new claim or proposal as what stands unless someone refutes it, as in the meeting example.
• Demanding the impossible. Asking for disproof of a claim built so that nothing could count against it.
• Evading the burden. Pragma-dialectics, an approach to argument that treats fallacies as broken rules of a good-faith discussion, lists evading the burden of proof (breaking the rule that “you must defend your standpoint if asked to do so”) and mis-allocating it among the moves that derail a discussion, according to the Stanford Encyclopedia of Philosophy‘s entry on fallacies. The burden can be evaded without shifting it to anyone, for instance by presenting a claim as too obvious to need support, or by repeating it.

The trick is old. C. L. Hamblin, in his history of fallacy theory, calls the attempt to shift the burden of proof to one’s opponent “a sophistical trick that Aristotle … does not mention,” noting that in the ancient Greek practice of dialectical debate the burden always lay with the questioner.

Related, but classified differently: the argument from ignorance. Sources don’t agree on how the two fit together. John Locke’s original description of the argumentum ad ignorantiam (1690), as quoted by Hamblin, is close to this move: requiring “the adversary to admit what they allege as a proof, or to assign a better.” Bradley Dowden’s list of fallacies in the Internet Encyclopedia of Philosophy describes the argument from ignorance as using “an unjustified attempt to shift the burden of proof,” but lists misplaced burden of proof as a separate fallacy. One way to see the difference: an Argument from ignorance draws a conclusion from the absence of proof (“nobody has disproven it, so it’s true”), while shifting the burden is a move in a discussion about who has to supply proof. They often come together: shifting the burden sets up the argument from ignorance that then claims the win.

Asking the other side to give a reason is often fair. It isn’t this fallacy when:

• An agreed rule sets the burden in advance. Courts, contests, safety procedures and organizations fix defaults for what happens when evidence runs out, often because one mistake is worse than the other. Treating a wire as live until it’s tested puts the burden on anyone who says it’s dead.
• A presumption backs the claim. Some claims start with a reasonable presumption in their favor: in Douglas Walton’s example from law, a properly addressed and stamped letter is presumed to have been received. Someone who says it wasn’t needs to give a reason. The best-known presumption, the presumption of innocence, works the same way: the accused doesn’t have to prove innocence, because the burden is placed on the prosecution (see the “not guilty” verdict in Argument from ignorance).
• The claim has already been supported. Once one side has given a reasonable case, the other side needs to answer it, not just repeat “I’m not convinced.” In law this is the burden of producing evidence, which Walton describes as shifting back and forth as the argument proceeds.
• The other person is the one making a claim. Someone who says “that’s false” or “that’s a fallacy” is asserting something too, and owes their own reasons.

The test: has the person making the claim given any reason for it, and is there an agreed rule or presumption that puts the burden on the other side?

“Live until you prove it’s dead”

An apprentice electrician says, “I switched the breaker off, so this wire’s dead.” The electrician replies, “It’s live until you’ve tested it. Show me the tester reading first.”

This sounds like “prove it isn’t”, and the electrician is demanding proof of a negative. But the default was set in advance, it applies to every wire, and it exists because the two possible mistakes aren’t equal: treating a dead wire as live costs a minute; treating a live wire as dead can kill. The test is also easy to do. That’s the agreed rule condition above.

The brake pads

A mechanic measures a customer’s brake pads, shows her the gauge reading 2 millimeters, and explains that the shop replaces pads below 3. The customer says, “Maybe they’re fine for a while longer.” The mechanic asks, “What makes you think so? The measurement says they’re worn.”

The mechanic is asking the customer for a reason, but only after giving his own: a measurement and a threshold. The customer is now the one making an unsupported claim (“they’re fine”), so asking her for support moves nothing that wasn’t already owed. That’s the claim has already been supported condition.

Asserting is easy and disproving is hard. Most claims can’t be conclusively refuted in the middle of a conversation, so asking the other person to refute yours is a cheap way to make it look strong, and their inability to answer on the spot feels like losing.

Outside a courtroom, nobody assigns burdens in advance. Each side can plausibly say the other owes the proof, and settling that question takes a separate discussion about the discussion. Walton, building on work by Erik Krabbe, models these disputes as metadialogues. Someone who simply announces where the burden lies, with confidence, often wins that discussion by default.

The courtroom model can also mislead. Ulrike Hahn and Mike Oaksford argue that in law the goal is to reach a practical decision, while in ordinary critical discussion the goal is often just to raise or lower how much a claim is believed, where crossing a threshold such as “reasonable doubt” may not matter. Seen that way, the useful question about “you haven’t disproven it” is not only who owed the proof, but whether anything has been said that should make anyone believe the claim more than before. Usually nothing has.

• Ask for the reason first. “Before we get to whether I can disprove it, what’s the reason to think it’s true?”
• Separate not believing from disbelieving. “I’m not saying it’s false. I’m saying I don’t have a reason to accept it yet.” That leaves the question open without taking on a burden you don’t have.
• Say what would count as support. A mark on the chair, a trial run of the tool: naming a concrete test turns a standoff into a way forward.
• If a rule really puts the burden on you, find out what the rule is. Sometimes the default is legitimate, and it’s better to meet it than to argue about it.