Thread by @DjangoWexler

Twitter: Hey, take this survey to help shape the future of Twitter!

Me: I guess I do have string opinions about that. *Clicks*

Survey: *doesn't work*

Seems about right?

STRONG opinions. The fact that this tweet has a typo is also peak Twitter.

Trying to think what my Twitter wishlist would actually be:

Easy -- spoiler button. Just like on reddit, blacked-out text that you have to click on to make visible. This seems like it would take about a day to implement.

Medium -- multiple "channels" for any given user account. Allow a user to separate cat pictures and political opinions and give followers the option to subscribe to either. Pure add-on, so can just be ignored and default is current behavior. Replaces clunky "mute my hashtag".

(Pro tip: "look for things users are kludging the platform into doing, and then implement that natively" is a pretty good strategy! It got us RTs, etc.)

Hard -- Real communities instead of feeds and hashtags, with fine-grained moderation controls. E.g. I can create a custom community, where only members I invite can post, and only people I invite can subscribe and see posts -- or it can be open, or whatever.

Not intrinsically hard, just seems like there'd be a lot of engineering. Basically the old LJ community model, but on Twitter.

There's the list @Twitter! @jack get on that when you get back from space or whatever.

Addendum -- Twitter has announced a new thing they're doing.

https://twitter.com/TwitterSupport/status/1417920469265788931

This is one of those ideas that seems good in abstract until it meets reality.

There's two ways you can go with this:

1) Train a filter for any individual person, so your up/down votes only apply to yourself. (That is, the "you" is singular.) This is dumb. It's just trying to train an algorithmic filter, when you could just have a robust manual filter.

2) Use it to train a *general* moderation algorithm (the "you" is plural) so that downvoted comments and others like them don't get shown to users.

The idea here is "well everyone hates some kinds of Twitter comments, so lets learn what those are and not show them!"

The problem is that different communities hate different kinds of comments.

The upvote-downvote feature works on Reddit (in the subs where it DOES work) because it's backed up by real moderation.

So a, say, liberal-politics community can upvote stuff they like, and downvote stuff they don't like, and it helps moderate -- but only because the membership can be policed by the moderators. People who violate the community rules can be kicked out.

Twitter has no equivalent. And so you'll have posts from one group retweeted for downvotes across another group, and brigade invasions to downvote things, and on and on.

Because -- say it with me -- ALGORITHMS CANNOT REPLACE HUMAN MODERATION. Not now and not in the foreseeable future. Natural language is haaaard.

Twitter is always chasing this fantasy of being a perfect neutral platform with mechanically-enforced rules because a) it's cheap, and b) it fits with the ideological predisposition of their leadership. But it's a fantasy. It's not possible.

Increasingly I think the closed-community route is the *only* possibility for reasonable internet moderation. The volume of traffic is just too high for paid moderators -- Twitter alone would need tens of thousands. You need to recruit volunteers, essentially.

Which is exactly what the reddit-style system does! And subreddits can be great or complete crap, depending on how good the mods are.

But, as I've said before: moderation is *not optional*. The first rule of internet communities is "any unmoderated space eventually becomes toxic."

If you are developing a social media thing and you haven't figured out how it is going to be moderated, that is as much of a failure of design as if the backing database doesn't work.