Yonatan Zunger  🔥

Twitter was so enamored of the idea that they had helped catalyze the Arab Spring that "free speech" became an unexamined article of faith.

Unexamined as in, whenever serious questions came up of "wait, does this actually help free speech?," the most naïve answer always won.

It's hard to think of a single case where Twitter's answer wasn't "allow everything, make it users' responsibility to block" —

Even when it was very clear that this imposed unscalable burdens on individual users, silenced *their* speech, or created public risks.

And Trump using Twitter was clearly far too exciting to leadership as well: "OMG we're right in the middle of the political process!"

The "public interest" exemption was largely shaped post-facto, in the same way that the lack of a hate speech rule was.

The fact is that this kind of speech drives traffic and press, and this counters investor concerns about lack of revenue.

I have had to sit and *make* these tradeoffs, so please don't try to bullshit me by explaining how it's more complicated than we think.

It is insanely complicated, one of the hardest things I've ever worked on, and I *still* know when I'm being bullshitted.

Twitter chose to optimize for traffic at the expense of user experience. That's why GamerGate, that's why Trump, that's why Nazis.

And Twitter's concept of itself as a "public forum" nonetheless shies away from the issues that every real public forum in the world sees.

If you have a vested interest in attracting speakers who draw the most traffic, you are not a neutral platform, and have to deal with that.

There is nothing at all wrong with that—few platforms *should* be neutral—but you can't act like it's not there.

If you're going to reap the benefits of having created one of the key sites where Nazis organize, you need to deal with the costs, too. //

I should spend some time being clear about what good solutions look like, too. This *can* be done properly.

I fully agree with the goal of maximizing speech. We benefit, as a society and as individuals, from a free marketplace of ideas.

But KEY POINT: People's speech can be used to suppress other people's speech. (Harassment, threats, etc)

Specifically, if someone can impose costs on another person for speaking, then speech becomes limited to those most able to pay those costs.

As Ian Gent's explanation of the Petrie Multiplier explains why simple blocking doesn't solve it: http://blog.ian.gent/2013/10/the-petrie-multiplier-why-attack-on.html …

The key result: Because speech can be used to suppress other speech, the speech maximum is *not* the zero-regulation point.

The zero-regulation point, "everyone can speak and can also block," means that someone targeted for mass harassment pays a much higher cost.

And differential costs for speech, especially when correlated with existing social differentials, mean you get nonuniform speech output.

And importantly, harassment breaks into "severe" and "mild" flavors, which are not uniformly distributed.

For example, while men and women experience ~ the same amount of total harassment, women experience ~2x the amount of severe harassment.

This is a key thing you need to understand when walking into a policy discussion.

This also ties policy to business goals: maximizing speech is related to maximizing engagement.

Doing this wrong will differentially reduce engagement by women, minorities, etc., which fundamentally narrows your advertising base.