This article is more than 1 year old
If you don't have anything nice to say, don't say anything at all: El Reg takes Twitter's anti-mean algorithm for a spin
We're nicer in person, we swear
The next time you craft an acerbic reply to a stranger, the Twitter iOS and Android apps will ask: "Want to review this before tweeting?"
Tweets are run through an algorithm designed to filter out "insults, strong language, or hateful remarks" with would-be broadcasters of hateful remarks prompted to edit or delete their post before sending it. Users can also opt to ignore Twitter's intervention and send the tweet in its original form.
According to Twitter, the feature has been in testing since last year, and has produced some measurable results. The company claimed that 34 per cent of those issued a warning opted to review or delete their reply. It also said that, once prompted, users sent 11 per cent fewer mean tweets going forwards.
Obviously, we're not going to take Twitter's word for it. And so, one fateful Thursday afternoon, your humble hack sent a call-out to his nearly 8,000 Twitter followers asking for volunteers willing to receive a verbal tongue lashing.
"Hey! I need to tweet mean things to people for an article I'm working on," I wrote. "If you don't mind me writing something slightly cruel to you, reply to this tweet."
Nearly 50 brave souls volunteered within the first hour. And thus commenced an amusing, albeit potentially career-limiting, experiment in stress-testing Twitter's content moderation algorithms.
All's fair in love and tweet-fare
Of course, it figures that the first people to reply were those I'd either worked for, worked with, or simply respect on a professional and personal level.
Kevin Raposo, editor of gadget site KnowTechie, stuck his head above the parapet, only to be informed that he "looks like Vanilla Ice, if Vanilla Ice had the Benjamin Button disease." I told journo Alex Wilhelm his writing was akin to taking Valium.
And then it was the turn of Boris Veldhuijzen Van Zanten, my former boss at Dutch tech blog The Next Web.
"You're such a massive dipshit, you should rename your company The Next Pleb," I said.
It was a strong start. Twitter said nothing, although I quickly realised that was my fault. I'd been using the web view, rather than the iOS app.
The experiment continued. I took out my phone and searched for my next victim. Eric McMahon, a hirsute church pastor from Philadelphia, was told his beard resembled something found in a discarded grumble mag from the 1970s. He did not reply.
I asked if redheaded games journo Liz Finnegan was embalmed, or just really pale. I instructed Drew Olanoff, another writer, to perform an obscene act with a mallard. I described PR people as "sociopaths" and "pointless".
The app said naught. Could it be that my insults were so esoteric and contrived, they simply sailed past Twitter’s content filters? Or could it be that I'm just so bad at internet smack talk, it simply deemed my insults as unworthy of intervention? Truthfully, dear reader, I have no idea.
But as someone who has been on Twitter for a long time, I can't fault the intention behind the move. It's a horrible, toxic, draining place. Social media in general has a tendency to make strangers feel remote and abstract. They aren't people, but rather pixels on a screen, and thus it's easier to tweet things you wouldn't dare say to their faces.
Of course, policing a site with hundreds of millions of users is a task of herculean proportions, and an algorithm can only go so far. The only cure is deleting the damn app from your phone.
The real world isn't perfect, certainly, but it's a bit nicer. ®