The Fine Art of Baloney Detection

In 1995, already ill with the disease that would kill him within the year, Carl Sagan published The Demon-Haunted World, a book about how not to be fooled. Its most quoted chapter hands you a set of tools he called a baloney detection kit: nine habits of mind for telling a real claim from a dressed-up one, plus a rogues' gallery of the roughly twenty ways an argument cheats. It reads like a checklist. That is exactly the problem with it.

A checklist is easy to admire and easy to forget. You nod at arguments from authority carry little weight, agree that a hypothesis ought to be falsifiable, and walk away unchanged, because none of it cost you anything to read. Sagan's kit was never mainly a stick for beating other people's nonsense. It was built, mostly, for the one person you cannot cross-examine: yourself. The tools that matter are the ones that catch you.

One of them bites harder than the rest, and it is the most boring on the list: if there is a chain of argument, every link in the chain must work — including the premise — not just most of them. It sounds like a truism. It is not. It is a piece of arithmetic, and the arithmetic runs against you in a way almost nobody feels coming.

A chain of strong links is a weak chain

Picture an argument as an actual chain. Each link is a step you are fairly sure of — say ninety percent, the kind of confidence that feels solid, the kind you would defend out loud. String seven of them together to reach your conclusion. How sure are you of the conclusion? Most people answer about ninety percent — as sure as the steps. The real answer is forty-eight. If the links are independent, their confidences multiply, and 0.9 multiplied by itself seven times is a coin flip. Seven strong steps, and the thing they prove is no surer than heads.

The chain above lets you feel it. Set how many links the argument has and how sure you are of each, and watch the conclusion — the red mark — settle far to the left of where it feels like it should. Add links and it slides further left. This is why a long, elaborate argument deserves more suspicion than a short one, not less: every extra step is another factor below one, another quiet subtraction from a total you were rounding up to certain.

Your brain does the multiplication backwards

There is a reason the answer surprises. In the early 1980s the psychologists Amos Tversky and Daniel Kahneman described a bug so reliable it earned a name: the conjunction fallacy. They sketched a woman — Linda, thirty-one, bright, outspoken, a onetime philosophy major concerned with discrimination and social justice — and asked which was more probable: that Linda is a bank teller, or that Linda is a bank teller and active in the feminist movement. Most people chose the second. It cannot be. Every feminist bank teller is already a bank teller; adding a condition can only shrink the odds, never grow them.

That is the same error wearing a different coat. Detail makes a story feel more probable, because detail makes it feel more real — more vivid, more coherent, easier to picture. But every added detail is another link, another number below one. The richly specified scenario that feels almost certain is, by construction, less likely than the vague one it was built from. We reward an argument for precisely the thing that should worry us: the abundance of moving parts that all have to be true at once.

The honest catch: when not to multiply

Now the part that keeps this tool from becoming its own kind of baloney. Multiplying assumes the links are independent — that each could fail on its own, for its own reasons. Real arguments are rarely like that. Their steps lean on a shared premise, so they tend to stand or fall together, and when they do, the chain is not the product of its links; it is closer to its single weakest one. So the honest reading of any argument is a range: no stronger than the weakest link, no weaker than all of them multiplied, and somewhere inside that band depending on how much the steps really move together. The tool draws you both ends. Where the truth sits between them is a judgment, not a sum.

An argument is only as strong as its weakest link — never as strong as its average one.

How to audit a chain

Sagan's kit is usually sold as protection from other people — the psychic, the salesman, the confident figure on the stage. It works there. But its nearer target is closer to home. The most fluent, best-decorated, most emotionally satisfying argument you will ever be asked to accept on thin evidence is one of your own, at two in the morning, about why the thing you fear is certain or the thing you want is sure to work. Every link feels like ninety percent. Do the multiplication anyway.

You can push the chain around yourself above — set the links, doubt one, watch the floor fall out of the conclusion. Then take it somewhere. The Belief Updater is the other half of the same discipline: moving a belief by the amount the evidence earns, and never all the way to certain. The Certainty Radius asks how far the damage travels when you sell other people more confidence than your chain can carry. And when two facts line up and a tidy story rushes in to connect them, The Coincidence Machine is the reminder that a chain can begin from a link that was never really there.