The Future Is Filters · Read Scott Alexander

Related to: The Toxoplasma of Rage

Tumblr Savior is a neat program that blocks Tumblr posts containing specific words or phrases. For example, if you don’t want to hear all of the excellent reasons going around Tumblr why you should kill all men, you just block “kill all men” and they never show up. Add a few extra terms like “white dudes” (nothing good ever came of an article including the phrase “white dudes”), “trans”, “cis”, and “pictures of my vagina”, and you can make Tumblr almost usable.

(My own Tumblr Savior list is an interesting record both of my psyche and of mid-2010s current events. Sometimes I imagine a future cyber-archaeologist stumbling across it and asking “But, but…why would he ban the word ‘puppies’?” Poor, poor innocent future archaeologist.)

I recently learned about Twitter blockbots. These are lists maintained by some trustworthy people, such that subscribing to the blockbot automatically blocks everyone on the list. The original was made by some people in the social justice community to help block people they figured other members of the social justice community wouldn’t want to have to deal with. Although some people seem to be added on by hand, the bot also makes educated guesses about who to block by blacklisting accounts that follow the feeds of too many anti-social-justice leaders.

There are rumors of a similar anti-SJ block list of people who engage on online mobbing and harassment in the name of social justice, but I can’t find it online right now and I think it might have been taken down.

An article I read recently (but which I can’t find right now to link to) proposes a higher-tech solution for Facebook’s harassment problems. They want Facebook to train machine-learning programs to detect posts that most people would consider trollish. So far, so boring. The interesting part comes afterwards – instead of auto-blocking those posts, Facebook would assign them a certain number of Troll Points. Users could then set an option for how their Facebook feed should react to Troll Points – for example, by blocking every post with more than a certain amount. That way, people who were concerned about free speech and who enjoy participating in “heated discussion” would be able to do so, while people who wanted a safer and more pleasant browsing experience could have a very low cutoff for taking action.

But the really interesting part got dismissed after a sentence. What if instead of combining everything into Troll Points, Facebook assigned the points in different domains? Foul Language, Blasphemy, Racial Slurs, Threats, Harassment, Dirty Argument Tactics, et cetera. And then I could set that I don’t care about Foul Language or Blasphemy, but I really don’t want to see any Threats or Racial Slurs.

(obviously the correct anarcho-capitalist solution is to have third-party companies making these algorithms and selling them to individual Facebook users, but in a world where Facebook is trying to become more and more closed to third-party apps, that’s probably not going to happen)

So, take all this filtering technology – Tumblr Savior, Twitter blockbots, and hypothetical Facebook Troll Points, combine them together, project them about ten years into the future with slightly better machine learning, and you have an Internet where nobody has to see, even for an instant, anything they don’t want to. What are the implications?

II.

The most obvious possibility is that everyone will be better off because we can avoid trolls. In this nice black-and-white worldview, there are good people, and there are trolls, and eliminating the trolls is a simple straightforward decision that makes the good people better off. This is how The Daily Beast thinks of it (How Block Bot Could Save The Internet), and as anyone who’s been trolled or harassed online knows, there’s a lot of truth to this view.

The second most obvious possibility is that we will become a civilization of wusses safely protected from ever having to hear an opinion we disagree with, or ever having our prejudices challenged. This is how Reason thinks of it (Block Bots Automate Epistemic Closure On Twitter). Surely there’s some truth here too. How hard would it be to create a filter that blocks all conservative/liberal opinions? Just guess based on whether a text links to foxnews.com or dailykos.com, or add in linguistic cues (“death tax”, “job creators”, etc). Once such a filter existed, how many people do you think would use it proudly, bragging about how they’re no longer “wasting their time listening patiently to bigots” or whatever?

But I don’t think the scenario is quite that apocalyptic. If you’re getting all of your exposure to opinions you disagree with from them being shouted in your face by people you can’t avoid, you probably are not going to lose much by not having that happen. The people who are actually interested in holding discussions can still do that. When I was young and therefore stupid I used to hang out at politics forums specifically for this purpose.

The third possibility is that there would be a remarkable shift of discourse in favor of the powerful and against the powerless.

Terrorism has always been a useful weapon of the powerless. The powerful get laws passed through Congress or whatever, but the powerless don’t have that opportunity. They need to get people to pay attention, and blowing those people up has always been an effective tool in that repertoire. We see this most obviously in places like Palestine and the Basque Country. Likewise, as many people have pointed out, the recent riots in Baltimore can be thought of as a group of powerless people trying to make their anger heard in one of the only ways available to them. It would be politically un-savvy to call this “terrorism”, but as acts of destruction intended to promote a political struggle, they probably fit into the same cluster.

But the next step down from terrorism is annoyism. Terrorism is meant to convince by terrorizing those who ignore your cause; annoyism is meant to convince by annoying people who ignore your cause. Think of a bunch of protesters shouting on a major road, or throwing red paint over people wearing fur, or passive-aggressive Tumblr posts starting “dear white dudes”, or, in probably the purest example of the idea, the Black Brunch protests, where a bunch of black people burst into predominantly white restaurants and shout at patrons about how they’re probably complicit with racism. Even if there’s no implicit threat of force, the point is it’s unpleasant and people can’t ignore it even if they want to.

And so the traditional revolutionary chant goes: “No justice, no peace.” But the thing about filters is that they offer the opportunity for peace regardless of whether or not there is justice. At least they do online, which is where people in the future are going to be spending a lot more of their time.

Imagine you are a rich person who doesn’t want to have to listen to people talking about how rich people need to be socially responsible all the time. It makes you feel guilty, and they are saying mean things like that you don’t deserve all of the money you have, and shouting about social parasites and so on.

So you tell your automated filter to just never let you see any message like that again.

There is an oft-discussed division between politically right or neutral loud angry people (“trolls”) and loud angry people on the political left, (“you are not allowed to dictate the terms on which victims of oppression express their righteous anger”). Machine learning programs will not accept that division, and the latter can be magicked out of visibility just as easily as the former.

Imagine being able to put an entire movement on mute. While I can’t deny the appeal, I’m not sure we – and especially not the social justice community, which is currently laughing at the complaints of people who object to their blockbot – have entirely thought this one through.

III.

The part I find most interesting about all of these possibilities is that they force us to bring previously unconscious social decisions into consciousness.

I think most people, if asked “Is it important to listen to arguments by people who disagree with you?” would answer in the affirmative. I also think most people don’t really do this. Maybe having to set a filter would make people explicitly choose to allow some contrary arguments in. Having done that, people could no longer complain about seeing them – they would feel more of an obligation to read and think about them. And of course, anyone looking for anything more than outrage-bait would choose to preferentially let in high-quality, non-insulting examples of disagreeing views, and so get inspired to think clearly instead of just starting one more rage spiral.

And I think most people, if asked “Is it important to listen to the concerns of the less powerful?” would also be pretty strongly in favor – with the caveat that people can recognize annoyism when it’s being used against them and aren’t especially tolerant of it. The ability to completely block out annoyism, combined with people being forced to explicitly choose to listen to alternative opinions, might make groups that currently favor annoyism change tactics to something more pleasant – though possibly less effective.

I think the result would be several carefully separated groups with their own social and epistemic norms, all of which coexist peacefully and in relative isolation from one another – groups which I would hope then develop their own norms about helping powerless members. This would be an interesting step towards what I describe in my Archipelago article as “a world where everyone is a member of more or less the community they deserve.”