The Twitter Moderation Idiocy

There's no good way to moderate content on a web site like Twitter.

A week or so ago on Twitter, Trans activists announced April 1 as a day of protests by trans people, curiously called the "Trans Day of Vengeance". Twitter didn't find this objectionable enough to moderate. (Or, more likely, this didn't come to the attention of any Twitter moderators or programs.) After a couple of days, a trans activist shot up a Christian school in Nashville, Tennessee. Soon after that, talk turned to how violent trans activists frequently are, and right wingers started tweeting about the TDoV announcement. Twitter people suddenly woke up from their slumber and decided to stop the spread of the announcement. Twitter employed a computer program to suppress it.

That moderation exercise was pointless - all the people that were supposed to get the news most likely already got it - but it effectively muzzled the right wingers. The computer program doing the moderation cannot tell if your tweet is meant to spread the news or criticize the trans goons. It is basically programmed to shut down any tweet that contains the TDoV announcement picture. It locked down many accounts, mostly those of right wingers critical of TDoV. So, basically, Twitter not only allowed the spread of the original TDoV announcement unfettered for several days, it stopped people from criticizing it. If Twitter wanted to lend a hand to the trans activists, it couldn't have planned its actions better!

This is usually what happens when a computer program 'moderates' content. It happened many times in the past on Twitter, under previous management. Unlike a human moderator, a program cannot discern. (And that is true even with AI, though this specific program appears to be a brute force enforcement without any AI. When people have trouble interpreting what people mean many times, we shouldn't expect computers to do that task well.) But a program is quick, and it scales well when dealing with a viral spread of a tweet. Still, given the possibility of error, Twitter shouldn't have used the program to lock accounts. Just deal with the tweet and leave the account alone.

A human moderator fares much better in identifying intent behind tweets. But a big problem with human moderators - apart from the slowness - is the baggage they bring to the task. For example, many on the left think speech can be violence. Though words can 'hurt' people's feelings (and cause 'psychic injury'), nobody in his right mind thinks words are violence. But we are talking about leftists here. So, to a leftist moderator, a tweet that says "men can never become women" may look like causing or promoting violence. If the Twitter rule is to suppress tweets promoting violence, then the tweet qualifies. This is not even a hypothetical example. This sort of thing used to happen all the time when Dorsey and Gulag Parag ran the show on Twitter.

The folks at The Federalist seem to think Twitter is broken beyond repair. Are the Twitter employees deliberately looking to suppress criticism of trans community, or are they just incompetent? Musk himself says all the right things on free speech, but he seems to think moderation is both desirable and possible. I didn't see him acknowledge how difficult it is to moderate properly. He also seems to let good people continue to suffer from "shadow banning" and other shady moderation practices at Twitter. Ella Irwin - the person in charge of "Trust and Safety" at Twitter - doesn't seem to appreciate how the automated scripts end up doing the wrong thing a lot of the times. I haven't seen her even acknowledge that accounts are getting locked up for wrong reasons now.

When Twitter turned up its moderation activities recently, it seems to have turned on some sort of sensor for 'violence'. I got caught in it and got locked out for a week. I made a joke about torturing Mayorkas, the bug eyed bum in charge of border enforcement under Bidentia. I post the tweet, and two seconds later, my account is locked. And Twitter says, "after careful review, we determined your account broke the Twitter rules." Careful review, my foot! A computer program did the deed. Again, how hard is it to delete the tweet without locking the account?

Links: