Against Lie Inflation · Read Scott Alexander

[Related to: The Whole City Is Center]

I got into an argument recently with somebody who used the word “lie” to refer to a person honestly reporting their unconsciously biased beliefs – her example was a tech entrepreneur so caught up in an atmosphere of hype that he makes absurdly optimistic predictions. I promised a post explaining why I don’t like that use of “lie”. This is that post.

A few months ago, a friend confessed that she had abused her boyfriend. I was shocked, because this friend is one of the kindest and gentlest people I know. I probed for details. She told me that sometimes she needed her boyfriend to do some favor for her, and he wouldn’t, so she would cry – not as an attempt to manipulate him, just because she was sad. She counted this as abuse, because her definition of “abuse” is “something that makes your partner feel bad about setting boundaries”. And when she cried, that made her boyfriend feel guilty about his boundary that he wasn’t going to do the favor.

We argued for a while about whether this was a good definition of abuse (it isn’t). But I had a bigger objection: this definition was so broad that everyone has committed abuse at some point.

My friend could have countered that this was a feature, not a bug. Standards have been (and should be) getting stricter. A thousand years ago, beating your wife wasn’t considered abuse as long as you didn’t maim her or something. A hundred years ago, you could bully and belittle someone all you wanted, but as long as there was no physical violence it wasn’t abuse. As society gets better and better at dealing with these issues, the definition of abuse gets broader. Maybe we should end up with a definition where basically everyone is an abuser.

But a wise supervillain once said, “When everyone is super, nobody is”. In the same way, when everyone is an abuser, nobody’s an abuser.

Right now, if I hear that someone is an serial abuser, I would be less likely to date them, or I might warn my friends away from them, or I might try not to support them socially. The world is divided into distinct categories – abuser and non-abuser – and which category someone is in gives you useful information about that person’s character. I’m not saying that every abuser is an awful person who is 100% defined by their misdeeds and can never be redeemed. But I think the category contains useful information about a person’s character and likely future actions.

But if everyone used my friend’s definition, and we acknowledged that everybody is an abuser – the category stops being informative. “John is an abuser”. So what? Doesn’t mean you should worry about John, doesn’t mean you shouldn’t date John, doesn’t even mean you shouldn’t set your single friends up on blind dates with John. It just means John is a human. Maybe he cries sometimes. So what?

Broadening the definition of “abuser” this far doesn’t help fight abuse or make anybody nicer. It just removes a useful word from the English language. I can still eventually warn someone that John is cruel or violent toward people close to him. I just have to circumlocute around the word “abuser”, in order to find some other word or phrase that hasn’t been rendered meaningless.

(I’m cheating here by talking about “abusers” rather than “abuse”, since there is still a useful distinction between abuse and non-abuse actions. But although the abuse case is less clear, I think some of the same considerations apply – just because an action is abuse no longer means you can be sure it’s especially bad)

But it’s worse than this, because change to a definition doesn’t instantaneously propagate to all of its web of connotations in our minds. So probably some people will continue to use the new definition while still holding the connotations of the old definition. This means bad actors can stigmatize anyone they want:

1. We don’t tolerate abusers around here, right? Right!
2. John’s actions technically qualify as abuse by this incredibly broad standard that includes basically everyone.
3. Therefore we shouldn’t tolerate John.

I previously called this manuever The Worst Argument In The World. Maybe that’s a slight exaggeration, I’m just really tired of seeing it again and again.

II.

This is also my objection to broadening the meaning of “lie”.

The word “lie” is useful because some statements are lies and others aren’t. And although people may disagree on which statements are lies or not (Did OJ lie when he said he was innocent? opinions differ!) everyone agrees on a mapping between states-of-the-world and lie-vs-truth status. When I say “OJ lied”, everyone understands me as making a specific claim about the world, which they can either accept or reject. I don’t have a lot of leeway in how I use the word “lie”; if I’m calling you a liar, I’m making a specific claim about the world.

If “lie” expands to include biased or motivated reasoning, who’s going to throw the first stone? We’re probably all biased to some degree. Does that make us all liars? If everyone’s a liar, nobody is. I can accuse Donald Trump of lying constantly, and you can just nod your head and say “Oh, so you’re saying he’s not a perfect person free from all bias, whatever”. You’ll feel no need to decrease your opinion of him.

Maybe we should only apply the word “lie” to particularly egregious bias and motivated reasoning? But we’ve already abandoned the only defensible Schelling fence. So how will people decide where to draw the line? My guess is: in a place drawn by bias and motivated reasoning, same way they decide everything else. The outgroup will be lying liars, and the ingroup will be decent people with ordinary human failings.

This is my criticism of the original post that started this argument. It declared belief in a near-term singularity to be a “scam” and that believers were “being duped into believing a lie”. Its evidence was listing some reasons people might be biased to believe the singularity was near.

The obvious next step is that someone who believes the singularity is near writes a post listing some biases that singularity skeptics probably hold (for example, the absurdity fallacy). Having shown that skeptics are biased, they pronounce skeptics to be liars perpetrating a massive fraud on the scientific community.

There are a few ways this expanded-definition world becomes different from the world where people restricted “lie” to mean a knowingly false statement.

First, everyone is much angrier. In the restricted-definition world, a few people write posts suggesting that there may be biases affecting the situation. In the expanded-definition world, those same people write posts accusing the other side of being liars perpetrating a fraud. I am willing to listen to people suggesting I might be biased, but if someone calls me a liar I’m going to be pretty angry and go into defensive mode. I’ll be less likely to hear them out and adjust my beliefs, and more likely to try to attack them.

Second, bad actors can use The Worst Argument In The World to prove whatever they want. As long as you’re willing to equivocate and deceive people, you can prove anyone a liar, and then draw on now-obsolete connotations of “liar” to silence or ostracize them.

Third, the biggest beneficiaries are actual liars. Suppose some singularitarian claims that internal Google documents prove they have already created human-level AI. And suppose that’s totally false and no such documents exist. Usually I would accuse them of lying, and this accusation would be enough to alert people that, hey, something has gone terribly wrong here. But if both sides are constantly accusing each other of lying just for having normal human failings, then “you are a liar” no longer carries much weight. I have to come up with some complex circumlocution in order to let people know that someone told a mistruth.

And that complex circumlocution can only last until people realize it too is an exploitable signal. The whole reason that rebranding lesser sins as “lying” is tempting is because everyone knows “lying” refers to something very bad. But the whole reason everyone knows “lying” refers to something very bad is because nobody has yet succeeded in rebranding it to mean lesser sins. The rebranding of lying is basically a parasitic process, exploiting the trust we have in a functioning piece of language until it’s lost all meaning – after which the parasitism will have to move on to whatever other trusted functional piece of language has sprung up to replace it.

III.

I realize this is a kind of long post arguing against a weird thing that not many people are doing. But I think it’s an especially clear case of a broader thing that many people are doing. Words like “disabled”, “queer”, and “autistic” are also gradually shifting meanings, getting applied more and more loosely.

This isn’t always bad! Words are useful because as they separate the world into categories; this suggests a word should apply in more than 0% of cases but less than 100% of cases. Exactly where it should fall in between that range, I don’t know. The broader you make the definition, the better the word’s ability to name things that have even a small level of the relevant quality. But the broader you make the definition, the less power the word will have to separate strong examples of a quality from marginal examples.

I think of this as a sort of sensitivity-and-specificity statistics problem, setting a threshold to divide the population into two groups. If you have a very strict threshold for “abuser”, maybe only someone who inflicts serious physical injuries, then you can use it to separate the most abusive 1% of people from the other 99%. If you have a very weak threshold for “abuser”, so low that 99% of people qualify, then you can use it to separate the 1% least abusive people from the other 99%. If you set it in the middle, you can separate the more abusive half of the population from the less abusive half. If “abuser” picks out the most abusive 1% of people, it transmits a lot of information in a small number of cases. If it picks out the most abusive 99% of people, it transmits very little information in a large number of cases (and now “not an abuser” transmits a large amount of information in a small number of cases!). If the boundary is set at 50%, it transmits an equal moderate amount of information about everyone.

There’s no rule that 50-50 is always the best – for example, if the word “murderer” referred to anyone in the more murderous half of the population, that would be much worse than the system now, where it refers to a much smaller set of people, who you have much more reason to worry about as a discrete group. You’re going to have to find the right threshold for each individual concept.

But it’s never the right decision to draw the line outside the population, so that literally 100% of people fall in one category and 0% in the other.

A few months ago I told a fable about a city. The citizens worried that people living in the outskirts of the city felt unimportant and excluded. So they redefined “city center” to mean the entire city, including the outskirts. Nobody ended up feeling any more important because of this, because living in city center stops being prestigious when everywhere is city center. But it was now impossible to direct tourists to where they wanted to go, and people had to invent new phrases like “the part of the city where there are the most tall buildings” in order to discuss city center.

The moral of the story is: don’t set thresholds for category membership so far outside a distribution that they stop conveying useful information.