In Defense of Twitter’s Decision to Ban Trump
After Twitter permanently suspended Donald Trump’s Twitter account, some high-profile rationalists and rationalist-adjacent folks came out strongly against the decision. Among those who seem strongly opposed: Eliezer Yudkowsky, Kevin Simler, Naval Ravikant, and Balaji Srinivasan.
This post is about why I think they’re wrong.
1. Does the law allow Twitter to do this?
Yes. Section 230 of the Communications Decency Act gives platforms broad discretion to edit or not edit user generated content as they see fit. Dozens of courts have held that this includes the ability to permanently ban users or suspend their accounts, for any reason or no reason. For more background on this, please read Professor Eric Goldman’s excellent Law and Technology Blog (Full disclosure: I’m an occasional contributor on the blog).
2. Does banning Trump stifle free speech?
No. Twitter is a private company and not a state actor. The First Amendment does not apply to decisions about whom it allows to use its platform.
Too many smart people are conflating the rights and responsibilities of state actors with those of private companies.
For example, in response to the ban, early-stage Twitter investor Naval Ravikant tweeted “First they ban accounts. Then, they ban apps. Finally, they block websites.”
It’s important that we disambiguate the “they” in this tweet. (Even though the misleadingly ambiguous pronoun choice gives the statements all of their purported power.)
There’s a critical difference between a state actor banning free speech and a private platform doing the same. The First Amendment provides the confines for government limitations on free speech. The law gives broad discretion to private companies to address what kind of speech to permit within the confines of their physical and digital boundaries.
Conflating the two is a lazy and misleading.
Donald Trump can still speak freely in public.
He still has plenty of soapboxes. Just not his preferred soapbox.
3. Is it wrong for Twitter to enforce social norms on its platform?
As a fan of SlateStarCodex, I remember the occasional post where Scott Alexander would give updates on commenters who received temporary or permanent bans for violating the rules and norms of his comments section. Occasionally, folks would kick and scream about those decisions. But Scott wanted to encourage good-faith arguments and civil and charitable conduct in his comments section. Scott understood that un-moderated comments sections often devolve into poisonous conversations. He policed the comments section as he saw fit.
I’m not intimately familiar with the content moderation policies at Less Wrong, but based on an initial Google search, it certainly appears that the powers that be have broad discretion to delete “anything [they] judge to be annoying or counterproductive.”
Why should Twitter be judged for doing the same thing? Donald Trump wasn’t removed from Twitter because of an ideological difference-of-opinion. He wasn’t banned for arguing for higher tariffs with China or restrictive immigration policies. He was suspended because his Twitter feed is a constant stream of poison and misinformation that threatens the entire ecosystem. This was true from the advent of birther-ism to his formal ban this week. He has, for months, publicly stated that he would not commit to a peaceful transfer of power. Now that it’s time to relinquish power, he incited a violent insurrection to stop the formal process of transitioning him out of power. If that’s not grounds for removal, what is?
If Twitter had banned Trump merely for taking an unpopular ideological stance, I would have opposed the ban. But Twitter banned for poisoning their ecosystem and repeatedly inciting violence.
4. Isn’t Twitter a monopoly and doesn’t banning him effectively prevent him from engaging in free speech?
No, Twitter is not a monopoly.
Only 22% of Americans have Twitter accounts.
Almost all Americans have TVs. Trump could give a TV press conference tomorrow and most networks would host it or at least publicize it.
And even if it were a monopoly, that wouldn’t affect the free-speech or Section 230 implications of this ban.
Again, Trump still has plenty of places to disseminate his poisonous message. Just not on Twitter.
5. But what about the slippery slope that will lead to the parade of horribles?
The law makes hard distinctions all of the time. Content moderation systems do, too. Twitter had to adjudicate the facts in front of it. That’s what it did here.
Donald Trump is an evil jackass and a legitimate threat to democracy. Purging him from the system is good for the system. He is actively trying to subvert a 230-year-old system of government. There are no parallels to this situation anywhere and there likely (hopefully) never will be again.
If Twitter makes another decision in a different context, I reserve the right to disagree with that decision. For example, I was against Google’s decision to fire James Damore for his amateur evo-psych missive.
But context matters in the world of content moderation.
6. Twitter’s rules are selectively enforced
Sure. But content moderation at scale is inordinately difficult. There will always be false positives and false negatives.
One of the most famous examples of content moderation gone bad is the famous picture of the naked child in Vietnam running from a Napalm explosion. A few years ago, Facebook’s algorithms flagged the picture as offensive and inserted black box over the child’s genitalia.
In almost every circumstance, it is wrong to post naked pictures of children. A rule against such posts would normally be a good thing. But in this instance, the consensus was that this was not the right decision.
Rationalists are inclined to seek out clear, bright-line rules. Unfortunately, content moderation does not lend itself to such clarity. Content moderation must involve ad hoc, fact-based adjudication, because context matters in content moderation. To quote the true expert on this subject, Mike Masnick:
Trump is, perhaps, the perfect example of why demanding clear rules on social media and how they moderate is stupid.
As for the question of why now? Well, clearly, the context has changed. The context is that Trump inspired a mob of goons to invade the Capitol building this week, and there remain legitimate threats that his cultish followers will continue to do significant damage. Certainly some people have insisted that this kind of violence was always a risk — and it was. But it had not actually erupted to this level in this fashion. Again, we’re talking about context. There’s always more context. And given that the situations are always edge cases, that the context always matters, and that things are always shifting, you can totally see why it’s a reasonable decision to ban Trump from their platforms right now, based on everything else going on, and the likelihood that he might inspire more violence.
If you really want to understand this issue, spend a few hours reading Mike Masnick’s historical posts on this subject.
Is Twitter’s current system imperfect and occasionally unfair? I’m sure it is. But like other imperfect systems, it’s the best we have. Well-intentioned policy-makers have been trying to come up with alternatives to Section 230 for years, but whenever such alternatives are subjected to careful scrutiny, most scholars on the subject conclude that the alternatives are far worse.
7. Will banning Trump and similar norm-violators from mainstream platforms lead to more violence and unrest?
This was an argument by Yudkowsky. I don’t think that’s true, either. What’s dangerous about Trump is that he has made norm-violation mainstream. Maybe a small percentage of wingnuts believed that prior elections were rigged, but these types of allegations from presidents, senators, and US Representatives en masse are a unique feature of the Trump administration. That’s led to a destabilization of the political infrastructure of this country.
There will always be wingnuts. What is unique about our current situation is how they have gone mainstream.
There’s plenty of evidence that the major platforms’ algorithms have contributed to conspiracy-minded thinking. That their content-moderation policies should attempt to offset some of these effects is not a bad thing.
There are always going to be services like 8chan where the norm-violators will congregate. But most people don’t want to spend their time on 8chan; they want to be on Facebook, where their friends and family are.
There will certainly be violence from norm-violators who aggregate on fringe sites in the future. But if we can take steps to reduce the influence of wingnuts in mainstream culture, we can perhaps limit future iterations of khakistocracy like the one we’ve been subjected to for the last four years.
8. Zuckerberg and Dorsey changed their minds on this issue for a reason
Facebook and Twitter didn’t want to be in this position. Both Zuckerberg and Dorsey both advocated for laissez faire principles of content moderation just a few years ago. But they soon realized that this position was naïve. They realized that to keep their own standing in the community, they needed to ban some norm-violators and moderate some content.
Zuckerberg and Dorsey now know that removing from the ecosystem those whose actions could destroy the ecosystem is a necessary precondition for a functioning ecosystem.
That’s the judgment call Twitter made here.
Donald Trump is a recidivist norm violator, whose norm violations have been as damaging to the United States as any in modern history. Twitter deemed it in the best interests of its ecosystem (and the broader ecosystem) to ban him.
So be it. We’ll all be better off for it.