Raymond's Weblog^

Free Speech is a Compression Problem

CW mention of several controversial topics

When is it reasonable to stop someone from conveying some message? Well, here are three obvious answers:

I think all of these answers are kind of wrong, but they all touch something real. My personal guess for the principled answer is something like:

Essentially the question I’m asking myself is: when someone else gets this message (which could be anything from 240 characters to 200 pages), and they try to work out what it means, and what the implications are, will they end up better understanding the world?

Because of this, I think it is basically always wrong to stop two adults from discussing something amongst themselves. I also think it is often fine to restrict people’s rights to say things in public even if those things are in some sense true, basically in proportion to how short the message is

Generalising ‘lying’

The easiest and most unambiguous form of lying is just things that contradict reality: the sky is red, I can see into the future, and the spirits say you have to give me your money. It’s pretty clear-cut, and it’s hard to do by accident.

But what about when someone just misrepresents something? Overemphasises a few details and omits a few others, without ever saying anything false? I think this is in practice most of what happens even in domains like politics. And it’s not even always on purpose - people have a habit of misrepresenting themselves without meaning to. But also, if you did want to mislead or manipulate people, my sense is that wholesale fabrication is harder and less effective than just massaging the truth, so in practice this is where even most of the intentional harm emerges.

Is this kind of behaviour ‘lying’? Well it’s clearly related. I think lying has maybe three components:

  1. You say a thing which will cause someone else to be misled about reality
  2. And the thing which you say is itself not drawn from reality
  3. And this is your intention

But the more general case of manipulation, misrepresentation, and so on, at least omits point 2, but I think it preserves point 1.

The key thing to notice here is that you don’t need point 2 to get point 1.

Supposing I’m describing to you a longstanding acquaintance of mine that you’ve never met, by giving you a few anecdotes about their behaviour. In the ideal world, I pick some representative examples of things they’ve done, effectively compressing my sense of them into a minute of conversation, and then you correctly decompress the examples into a holistic sense of the kind of person that would have done those things.

But if I wanted, instead of representative anecdotes, I could pick only the most positive, or the most negative. Without lying, I could set you up with data such that, when you try to decompress it into the holistic sense of the person, you come out thinking they’re rather more kind, mean, competent, lazy, or whatever else. Notably, I probably couldn’t get you to believe that they were unusually violent or something like that, but I could shift your sense of how they fell within the normal range of people.

So it ends up being possible to mislead someone about reality without ever saying something false, but in this case, it requires that the someone do some decompressing in their head.

(I also note that there’s a whole other kettle of fish here about saying true things to people who already have very inaccurate beliefs. More on that in the section after next.)

Battles over categories are battles over natural decompression

If you haven’t already read The Noncentral Fallacy, I recommend you pause now to look at it - Scott is an excellent writer and this is one of his best essays.

In brief, he gestures at the argumentative pattern ‘X is a Y; the prototypical Y is Z; so X is Z’. For instance: ‘MLK is a criminal; the prototypical criminal is bad and should not be esteemed; so MLK is bad and should not be esteemed’. The second two parts are left implicit. But once you see it, it’s everywhere!

I would go so far as to say the majority of widespread cultural battles are about categories and compression: Is abortion murder? Is Hamas a terrorist group? Are trans women women? Is race genetically determined?

The reason these things become the focus of the debate is that that’s all there’s room for, but it’s understood that the resolution of this question will bear on hard implications. We don’t have cultural bandwidth for nuanced discussions of the fuzziness of all categories, so unfortunately there is a strong pressure to round things off entirely. Either abortion is 100% murder or 0% murder.

And it’s crucial to notice that the stakes and damage aren’t coming in the question itself, they’re coming in the implications.

Forbidden knowledge

Isn’t it funny how often historical consensus is laughably wrong? How bizarre to think people really believed that Earth was the centre of the universe, or that animals evolve a little over the course of their life and pass it down to their children. Good thing we dodged that bullet.

But these beliefs never exist in a vacuum. The USSR’s support of Darwinism, and later its firm rejection, were not arbitrary. What mattered was how Darwinism fit into a wider model of the world which, crucially, needed to support their ideology.

So it is, I think, uncontroversial to say that there are obviously true things which you ‘are not allowed to say’. But they’re less like random landmines. The issue is that they sit close in the web of belief to very dear matters of ideology and culture.

As I write this, I feel a little faint of heart, because I’m not actually sure how to do justice to the things I want to say next. Perhaps I’ll come back and do that later, or perhaps not. But if you’re reading this, I suppose you’d better try to fill in the blanks.