Free Speech and Hate Speech
Why “apolitical” or “symmetric” laws aren’t.
(This was earlier posted as a Twitter rant, here formatted nicely for your sanity)
Zoé Samudzi said something important today:
This was a response to the argument that bans on hate speech violate the basic principles of free speech, and that doing so threatens democracy itself. I subscribed to that idea myself until a few years ago, when I had to work on real problems involving speech policy; what I learned is that Samudzi is completely right.
Hate speech and harassment, if either de jure or de facto legal, silence the weak and amplify the powerful.
Hate speech and harassment (which is the personal, rather than group-focused, cousin of hate speech) impose costs on others for speaking. In so doing, they limit speech to those most able to pay those costs. This is the simple secret behind harassment campaigns, especially ones on the Internet. It’s also why “real name” policies have no effect on abuse: the offenders generally get social credit in their communities for doing so. (There are plenty of communities in which being an asshole not only costs you nothing, it’s a way to show off. Vide Trump.)
On the other hand, hate and harassment can have costs for the target, ranging from the simple work of having to plough through it, up through massive personal and professional consequences (Danielle Citron’s book Hate Crimes in Cyberspace is the best current reference on how). The asymmetries come from all sorts of sources: existing social asymmetries, e.g. women and people of color already having to “prove themselves” more to get the same job, and thus being more damaged by false statements online; personal asymmetries, like health or family issues; but most of all things intrinsic to harassment, like how thousands of people can do individually insignificant things that have a massive cumulative effect on one person (see also the “Petrie Multiplier,” a tremendously important idea in social networks).
This isn’t even just a moral argument: it goes directly to the idea that the purpose of the First Amendment is to create a “marketplace of ideas.”
All of this goes to Samudzi’s original point: laws which are “blind” to asymmetries which exist in the world have asymmetric results. This isn’t even just a moral argument: it goes directly to the idea that the purpose of the First Amendment is to create a “marketplace of ideas.”
The First Amendment was created with a specific, deadly threat to this marketplace in mind: the state using its power to pick and choose ideas. But to create a thriving marketplace, we need to ensure that other methods of suppressing speech don’t destroy it, either. In particular, people and groups using speech to prevent other people’s speech can close the marketplace just as thoroughly. This is especially important when the shutdown of speech isn’t politically uniform — if it shuts down some speech more than others.
Hate speech has the particular ability to shut down speech by minority groups more than that by majority groups. The powerful can shrug it off; others can’t. It’s not that it destroys people, but it imposes a cost on their speech or their visibility. And costs are what keep people out of a market. In this way, hate speech harms the very marketplace of ideas that is fundamental to our democracy and our society.
A counterargument to what I just said is that “the remedy to speech is more speech.” But against hate and harassment, this argument has two flaws. The first is the same reason that your ability to sue a bank isn’t usually a balance to the bank’s ability to charge you bogus fees: asymmetric resources. You both have to pay to have this adversarial contest, so whoever has the most resources can grind the other down. (“Millstone them,” for those of you who played Magic: The Gathering back in the day…)
The second is even simpler: it’s because there’s no meaningful argument to have around hate speech.
“I think taxes should be raised to pay for X” / “No, that’s a waste of money” is an important political debate. “Victimless crimes should be abolished” / “No, they have hidden victims” can be an important debate. But “You are subhuman and nothing you say has any meaning, because of your intrinsic foolishness” isn’t part of any debate. “I think you and your family should be killed.” / “I disagree!” is definitely not a debate. No new knowledge comes out of it.
And neither the person forced to “defend” their humanity, nor any of the bystanders who learn from real arguments, benefit from it being treated as one. The only thing this “debate” does is legitimize the idea that maybe some group is subhuman, because people evidently are unsure enough to discuss it. Legitimizing the dehumanization of others as a debate benefits one group only: those who wish to dehumanize. As a result, in such a conversation, it doesn’t even matter what they say; the arguments can be utter nonsense or change every minute. They get their desired benefit from the simple fact that the conversation is being had, and being granted legitimacy. There is no meaningful speech remedy to that.
Legitimizing the dehumanization of others as a debate benefits one group only: those who wish to dehumanize.
The net result is that hate speech harms the marketplace of ideas, by imposing costs on some speakers but not others, and it cannot be effectively stopped by the internal functions of that marketplace. Which is to say, it is a type of market failure, and so can only be remedied through means outside that market: regulation.
This does, of course, require a profoundly delicate balance: the threat to speech which state intervention poses (or really, any force majeure, not just the state’s) is real and profound. But this implies that the point which maximizes the freedom of ideas is not the zero-regulation extreme — which favors the ideas of the powerful — but some type of balance.
The Devil is, of course, in the details.
Devising good speech policy is profoundly hard, because the range of speech is incredibly nuanced. It’s not as simple as “bad words.” To someone who thinks we could, at the least, ban “really bad” words like n — , try applying that policy to:
(1) Langston Hughes’ poem “Christ in Alabama” (with thanks to James H. Cone’s The Cross and the Lynching Tree for teaching me the profound significance and political importance of Hughes’ metaphor);
(2) This routine by Chris Rock;
(3) That same routine, if he had done it for a white audience;
(4) Or that same routine, if it were instead repeated verbatim by Ted Nugent for one of his audiences.
Speech involves a lot of context and is hard to simplify. But one useful rule is this: the gentler the punishments for breaking a rule, and the wider the spectrum of such available, the easier it is to write rules that deal with fuzzy corner cases without catastrophic results. If you only have harsh punishments available, you need to set the bar for them extremely high. And if the law is not a “bright line,” as speech law can’t be, then people will shy away from its boundary — the classic chilling effect. For the state, whose gentlest punishment already involves the courts, that means there’s good reason to set a high bar and allow a lot. For a platform whose strongest sanction is to kick someone out, stronger rules are both possible and desirable. (Not least because, unlike the state’s rules, those rules apply at most to one forum, not to all fora.)
But even the state has reason not to be absolutely permissive, for the same reason that the First Amendment doesn’t protect extortion or fraud — even though those, too, are “speech acts.” If the purpose of the law is to maximize the marketplace of ideas — that basis so crucial to democracy and freedom — then the law must wrestle with this hard problem, and try to place a meaningful dividing line.
Hate speech and harassment, if either de jure or de facto legal, silence the weak and amplify the powerful. Thus seemingly “apolitical” or “symmetric” treatments of speech have profoundly political consequences, and we must never pretend that they don’t.
