Twitter sees 829,094,400 tweets/day. Wikipedia has 166,666 edits/day. So we're already talking about a factor of nearly 5000x. The number of contributors is also spectacularly different: according to [1] Wikipedia had edits from only 122,821 distinct users in the last 30 days. Twitter claims 237.8 million mDAU (which could be a low or high count, since the "m" stands for monetizable and users may not post.)

[1] https://en.wikipedia.org/wiki/Wikipedia:Who_writes_Wikipedia...

But only a tiny fraction of the 829 M tweets a day are abusive.

> But only a tiny fraction of the 829 M tweets a day are abusive.

But you have to read all 829 M to find the abusive ones—and you have to agree on what 'abusive' means.

I don't think the automated abuse detection systems need to go away. They should still be around, but only serving to notify human moderators of what they find, not doing any deletions or bans themselves.

Rely on users to report the bad apples, ban accounts who abuse the report features, and provide people whose identity has been proven with a government-issued ID card with a safety switch that enforces human moderation for "spam" reports.

Right now Twitter isn't doing anything but the first part, which is why "I got mass report banned on Twitter" stories are so commonplace. They need not be, if Twitter would do their goddamn job.

A major financial criticism of Twitter is that their profits are too low compared to their revenues, mostly because of high salaries. Objectively, they're "doing their goddamn job" too well right now: I doubt a newly debt-laden private Twitter will be able to afford an increase in these human resources.

Why would you need to do that? You can just have humans review the reports that algorithms generate...

> Why would you need to do that? You can just have humans review the reports that algorithms generate...

At some point, in designing the algorithms, in instructing the human reviewers, and so on, you still need to agree on what abusive means. So there's no way around that.

As to the volume, I agree, one can deploy algorithms as a first filtering step. In this case, I don't usually trust humans to be smarter than algorithms, but I do trust humans to be better at abuse than algorithms are at recognising it—unless the algorithms have so many false positives that they're barely filtering at all.

Let the users curate their own feeds rather than trying to curate things centrally. If some threshold of users have blocked one twitter user, then ban them maybe, otherwise leave it alone. They can remove obvious spam or whatever, but let me choose whether to ban that particular take on a subject or not from my feed.

I'm likely in the minority on this, but I'd rather have the control than trust others to moderate in my best interests for me.

> Haven't Wikipedia and Stack Overflow figured out the right way to moderate at scale?

I think their different goals make it easier (which is not to say easy!) to figure out an appropriate moderation style. Wikipedia is an encyclopedia; it is easier to impose norms on an encyclopedia than a social platform. I don't have such a pithy summary of what SO is, but, as basically a knowledge repository rather than a social platform, it, too, can afford to impose certain norms that would stifle a social space. And yet, even with this head start, they can still face considerable flare-ups of controversy!