Pixel Envy

Written by Nick Heer.

Twitter’s Algorithmic Timeline Sometimes Surfaces Radical Tweets, Boosting Their Reach

Oliver Darcy, CNN:

Over the last several months, Twitter has begun inserting what it believes to be relevant and popular tweets into the feeds of people who do not subscribe to the accounts that posted them. In other words, Twitter has started showing users tweets from accounts that are followed by those they follow. This practice is different from the promoted content paid for by advertisers, as Twitter is putting these posts into the feeds of users without being paid and without consent from users.

[…]

In effect, the practice means Twitter may at times end up amplifying inflammatory political rhetoric, misinformation, conspiracy theories, and flat out lies to its users. This comes at a time when other platforms, like YouTube, are facing intense criticism for using algorithms to suggest content to users. It’s been documented, for instance, that YouTube’s algorithm has exposed users to fringe content and helped radicalize them online. YouTube has pledged to address the problem.

[…]

There is some irony to the amplification of these right-wing voices. Trump and other prominent Republicans have long accused Twitter of “shadow banning” users with conservative viewpoints, an accusation Twitter has strongly denied. In reality, not only is Twitter not “shadow banning” these right-wing personalities for their political viewpoints, the platform’s algorithm is actually amplifying some of their tweets to audiences who do not even follow their accounts.

Algorithmic recommendations from all major platforms — Instagram and YouTube, especially — are failing users by encouraging them to take their interests to the farthest maximum: a you like coffee; have you tried cocaine? kind of effect. Are these features necessary? I imagine recommendations increase the amount of time spent on these platforms which, in turn, increases their ad revenue and improves the figures they report every quarter. Evidence is growing, though, that recommendations are also detrimental to the health of the platform and its users.

I hate to sound like a Luddite, but was there something wrong with a purely reverse-chronological feed?