Middle ground fallacy

Splitting a fact

Ana: The museum opens at 9 on Saturdays.

Ben: No, it’s 11.

Cam: You’re both sure, so let’s meet there at 10. That’s probably about right.

The opening time is a fact that one of them may have right. Averaging their claims doesn’t make either more accurate; if Ana is right they’ll be an hour late, and if Ben is right they’ll wait an hour outside. Nothing about 10 o’clock is supported by any evidence at all. This is the clear-cut case: a question with a checkable answer, settled by splitting the difference.

The onlooker

A neighborhood forum thread is arguing about the big oak on Elm Street. A certified arborist who inspected it posts that the tree is healthy. Another resident posts that it’s rotten inside and will come down this winter. A third reader, who hasn’t seen the tree, sums up: “The truth is probably in between. It’s likely somewhat unhealthy and should be cut back hard.”

The reader isn’t a party to the dispute, but treats the dispute itself as evidence about the tree. It isn’t: one post comes from an inspection, the other from a guess, and the middle position has no support of its own. If the second resident had claimed the tree would fall tomorrow, the same reasoning would have produced a gloomier “middle”. Scott Aikin and John Casey describe this pattern as treating the mere existence of a controversy as evidence about the state of the evidence.

A compromise reported as a finding

Two roommates argue over who broke the blender. Each says the other left it running with the lid off. A third roommate, tired of the argument, declares: “You’re probably both partly to blame. Split the cost of a new one.” A week later she tells a friend, “It turned out they were both partly responsible.”

Splitting the cost may be a sensible way to end the argument, since nobody can prove what happened. The error is in the last sentence: a settlement reached so everyone could move on has been retold as a discovery about who broke the blender. Jan Albert van Laar and Erik Krabbe call this the core of the fallacy of middle ground: mistaking a compromise, which settles a disagreement, for a resolution, which answers it. They note that a quarrel is sometimes taken as evidence that both parties must be to blame, which is this move.

The argument, spelled out:

• A says X. B says Y.
• Therefore the truth lies between X and Y.

It needs an unstated premise, the truth lies between the positions people take, which is not generally true. Laid next to the legitimate practice it resembles:

It sits roughly opposite the false dilemma. A false dilemma hides the options between two positions and insists on one of them; the middle ground fallacy insists on the options between and rules out both ends. Both let the menu of positions, rather than the evidence, decide.

• Splitting the difference on a fact: averaging two claims about a date, a number or a measure when the claims aren’t independent estimates of it, as in the museum example.
• The onlooker’s inference: a third party concludes the truth must be in the middle. Aikin and Casey quote Grant Sterling’s definition in the fallacy reference Bad Arguments: two or more people have presented conflicting views, and a third party assumes, without offering reasons, that the truth lies between the extremes.
• Compromise mistaken for resolution: parties who have negotiated a middle position present it, to themselves or others, as what the evidence showed. This is the first variant van Laar and Krabbe describe.
• “No middle ground”: the second variant van Laar and Krabbe describe runs the other way: refusing any compromise on the grounds that compromising would mean giving up one’s view, when a compromise settles what to do without requiring anyone to change their mind. (The phrase “no middle ground” is also used for the false dilemma, a different error.)
• Bothsiderism: Aikin and Casey’s name for a wider family in which a persistent disagreement is taken as a reason for some kind of moderation: splitting the difference, suspending judgment, giving “both sides” equal time, or refusing to take a side at all. They call it a meta-argumentative fallacy, an error in reasoning about arguments (the state of a debate) rather than about the subject of the debate. They trace an early version to Richard Whately’s Elements of Logic, where people who object to a reform urge onlookers merely to suspend judgment while “there are reasons on both sides”. Since there always will be, Whately wrote, that suspension “is practically the very same thing as a decision in favour of the existing state of things. ‘Not to resolve, is to resolve.’” (Compare Status quo bias.) Aikin and Casey also compare bothsiderism to the Argument from ignorance: both shift the burden of proof, here onto whoever claims the question is settled.

Names. The error goes by many names, which Aikin and Casey list: middle ground, moderation, false compromise, golden mean and gray fallacy, among others. The Latin argumentum ad temperantiam circulates too, but Aikin and Casey report that they could find no source for it; like many Latin fallacy names, it appears to be a modern coinage rather than a classical one.

A middle position is often the right one. What matters is whether anything other than its location supports it.

• The question is what to do, not what’s true. Van Laar and Krabbe argue that shifting from arguing to negotiating can be entirely reasonable, for instance when a decision is needed and neither side can persuade the other. A compromise settles a disagreement without resolving it, and there’s nothing wrong with it as long as nobody presents it as proof that the middle was right.
• The evidence supports an intermediate answer. Sometimes the facts really are in between: a cause with several contributors, or a plan with real costs and real benefits. The middle is then a conclusion from evidence, and it would be just as justified if nobody had argued for the extremes.
• The positions are independent estimates of the same quantity. When two honest measurements or estimates differ, combining them can be a sensible way to reduce error. The reason is the evidence behind each, not the fact that they disagree.
• Suspending judgment is warranted by the evidence. Aikin and Casey note that moderation has legitimate uses when there is evidence that the disagreement is genuine and the question uncertain. The error is treating any disagreement, however lopsided, that way.

The test: would the middle position still look right if the extreme positions had been different?

Setting the thermostat

Two housemates can’t agree on the thermostat. One is comfortable at 68°F, the other at 74°F. After some back and forth, they settle on 71 for the shared rooms, and each adjusts their own bedroom.

Neither housemate thinks 71 is the objectively correct temperature, and neither claims the argument proved it. They’re deciding what to do about a difference in preferences, and each prefers the deal to arguing every evening. This is the what to do, not what’s true condition above: a negotiated compromise, openly presented as one.

Two tape measures

Measuring a room for a rug, Kofi gets 12 feet 3 inches with one tape measure and his partner gets 12 feet 5 inches with another. They use 12 feet 4 inches, since neither tape is obviously more accurate.

This splits the difference between two readings, but the readings aren’t positions in an argument; they’re independent attempts to measure the same length, each with a little error. Averaging them is a reasonable way to combine that evidence. This is the independent estimates condition: the middle is supported by the measurements, not by the disagreement. If one tape were known to be stretched, the reasoning would change, and they’d use the other.

Moderation looks like fairness and open-mindedness, and splitting the difference ends an argument without anyone having to lose. Van Laar and Krabbe describe people being “beguiled by the popular saying that the truth must be somewhere in the middle”.

Aikin and Casey suggest a heuristic behind it: people in an argument tend to overstate their own side, so an onlooker without expertise discounts both claims and splits the difference. That can go wrong whenever one side hasn’t exaggerated. They also point to widely shared norms of balance and of hearing every side, which a weak position can exploit: merely continuing to disagree, with no new evidence, can be enough to be treated as one of “two sides”. The result, they argue, is an outsized advantage for weak views.

The fallacy also rewards extremes. If the truth is expected to be in the middle, anyone who wants to move the middle only has to take a more extreme position.

• Ask what supports the middle position itself. “What evidence points to 10 o’clock, other than that it’s between 9 and 11?”
• Separate “what’s true?” from “what shall we do?” A compromise can be the right decision without being the right answer. Say which one you’re offering.
• Look at the reasons, not just the positions. Two claims aren’t equal because there are two of them. Ask what each is based on.
• Check how the endpoints were chosen. Would the “middle” change if someone had argued something more extreme?
• Don’t swing to the other extreme. The middle position may still be true; the disagreement just isn’t evidence for it.