Why is a lie so much cheaper than the truth?

Picture an evening in January 2013. A software developer named Alberto Brandolini is off the clock, partway through Daniel Kahneman’s Thinking, Fast and Slow, television running low in the background. The channel has landed on one of Italy’s political talk shows, former prime minister Silvio Berlusconi on one side, the journalist Marco Travaglio on the other. Brandolini watches for a while. Then he picks up his phone and, on 11 January 2013, posts one sentence that a decade later still gets forwarded as though someone just thought of it this morning.

Here is exactly what he typed. The bullshit asimmetry: the amount of energy needed to refute bullshit is an order of magnitude bigger than to produce it. Notice the misspelling, asimmetry for asymmetry, typed fast while two men argued on the screen behind him. That typo turns out to be useful: three separate copies of the tweet, pulled independently years apart, all carry the identical slip, about as close to a fingerprint as a sentence gets.

The version most people have seen by now is smoother: The amount of energy needed to refute bullshit is an order of magnitude bigger than that needed to produce it. Cleaner grammar, half a step removed from what he actually typed. Use the rough original here. The mess is part of what makes it verifiable.

Sit with what each side of that sentence actually requires, because the gap is a checklist, plain and countable.

Making a false claim costs one sentence and a confident tone. Nothing needs sourcing. Nothing needs checking against a second account. You can invent a refund policy, a lab result, a company’s founding date, a stranger’s motive, and the only price is the two seconds it takes to type it.

Taking the claim apart costs a search, at minimum. Usually it costs several: finding where the claim came from, finding what actually happened instead, holding both up against each other, writing the correction so a stranger can follow it in under a minute, and then, the expensive part, getting that stranger to trade the version they already absorbed for the version that took you an afternoon to build. Every one of those steps is optional for the person lying and mandatory for the person checking.

Brandolini put a number on that gap: an order of magnitude, roughly ten times. Take the figure for what it is, a gut estimate from a man half-watching television, typed in one sitting with no citation behind it. A tweet is a frame, sharp and portable. It is a different kind of object from a measurement, and the measurement, when it eventually arrived, told its own separate story, worth its own section.

Here is that arithmetic, drawn as a small room instead of a sentence. Ten dots, wired the way people actually cluster: a few close connections near the center, and past them, people who only know a friend of a friend. Drop one claim on the center dot and watch where it travels.

The claim reaches every dot in the room before the correction has covered half of it. By the time the correction finally rings the nodes closest to the center, the claim already stands two hops further out, filled in, finished, and the outer ring never gets a ring of its own at all. Read the two bars beside the network the same way. One rises and stops almost before you notice it moving. The other keeps climbing for as long as you watch it: cheap to make, expensive to unmake, exactly the shape of the sentence a programmer typed in 2013.

Five years after that tweet, in March 2018, three researchers, Soroush Vosoughi, Deb Roy and Sinan Aral, published the closest thing this subject has to a real measurement. Their journal was Science; their dataset was the whole slice of Twitter they could reconstruct: roughly 126,000 cascades of news stories, tweeted a cumulative 4.5 million times by about 3 million people, over a decade, 2006 to 2017.

False stories were 70 percent more likely to be retweeted than true ones. A true story took about six times as long as a false one to reach 1,500 people. A false story reached ten retweets deep, what the researchers called a cascade, roughly twenty times faster than a true one reached the same depth. Their own summary line: falsehood diffused significantly farther, faster, deeper, and more broadly than the truth in all categories of information.

Keep that finding in its own lane. Brandolini was describing effort, the labor of building a lie against the labor of dismantling one. Vosoughi, Roy and Aral measured something else entirely: speed and reach, on one network, across one specific decade. The two ideas happen to share a phrase. MIT’s own writeup of the paper used the words by an order of magnitude to describe how far ahead falsehood ran in some categories, and that shared phrase is where the resemblance ends: two separate claims, each carrying its own evidence.

The researchers had their own explanation for the speed, more useful than any raw ratio: falsehood scored higher on novelty in their data, and people share what feels new. A lie is cheap to make surprising, since none of it has to survive contact with a fact-checker before it ships. The truth stays stuck being whatever actually happened, which is only occasionally shocking. Jonathan Swift noticed the shape of this three centuries early, writing from inside eighteenth-century London politics on 9 November 1710: Falsehood flies, and truth comes limping after it. The line gets handed to Twain or to Churchill more often than it gets handed to Swift. A 2017 PolitiFact check traced a Churchill version back to the same 1710 essay and rated the attribution False. Swift wrote it. Give him the credit.

Here is where the arithmetic stops staying abstract. If refuting costs an order of magnitude more than lying, and a correction starts late and travels slow by nature, answering every claim by hand, one at a time, forever, is a race built to lose. The only move that changes the math is refusing to run the same check twice. PayPal’s own promise, that refunds take 3 to 5 business days, got checked once, on 9 July 2026: fifty reports, eight platforms, and a finding of 2 to 10 days, often longer. That work happened a single time and stands, signed, for the next person who wonders the same thing.