The trouble with algorithmic feeds, on social media or YouTube or TikTok or wherever, is that they put a thumb on the scale for bad behavior. That's, at least, the folk diagnosis. And so the folk remedy, to most if not all of the problems of social media, is to replace algorithmic feeds with strictly chronological ones. Either by having customers demand it or, if that fails, having governments dictate it.

I'm skeptical of this folk theory on both its diagnosis and remedy sides. Skeptical in that, if you were to get most or all social media platforms to stop filtering what you see based on what their machine learning tools think you'll engage with, and instead had them just show you everything from everyone you follow, with the latest at the top, I'm pretty sure it wouldn't actually change all that much. The problems with how people use social media—and what social media does to our information environment, our culture, and our politics—are complex and deep and not easily solved by a "flip this switch" silver bullet.

Roughly, here's what purported to be wrong with algorithmic feeds, and what, it's believed, chronological feeds will fix.

• Gaming the algorithm. The basic motivator of bad behavior is a desire for engagement. We want lots of likes, reposts, followers, clicks on our links, etc. If the algorithm is what determines how many people see each of our posts, we'll write our posts with the algorithm's preferences in mind, instead of aiming for our posts to have free-standing value that contributes positively to the public conversation.

• Rage farming. Gaming the algorithm leads, frequently, to rage farming. Platforms want users to maximize their time on site, which means maximizing their engagement, which means showing them posts they're likely to engage with, and one of the best ways to get people to engage with a post is to show them posts that'll piss them off. Hate reposts are still reposts. Quote dunking is still quoting. The more people are enraged by what they see, the more they'll stick around to keep that feeling going, or to make sure they don't miss something they ought to get mad about. This is Fox News's strategy for keeping parents and grandparents glued to their TVs, and it works.

• Fickleness. The first two are mostly problems for all of us as users of social media. It's not psychologically good to be in a place where everything you see is tuned to upset you. The problem of fickleness is more a concern for content creators. If you want to build an audience on a social media platform, you need to know what kind of content will grow that audience. And you want to know that if you commit to a strategy now, it won't blow up on you tomorrow. If the powers that be retune the algorithm, however, what works today might not work tomorrow, and you'll lose all your algorithmic reach. As a creator, then, algorithmic feeds can lead to your livelihood getting burned.

• A thumb on the scale for big/high engagement accounts. This is about creators, too, but primarily smaller ones. If the algorithm optimizes on engagement, and part of that engagement is that people like to see content from big accounts, it can be hard to become a big account if you aren't a first mover on the platform, or aren't independently famous. And until an algorithm knows enough about a new user to have a full sense of their interests, it's likely to lean on just showing them the most popular stuff, which is likely to be big accounts. So algorithms can make it harder for new people to build a following.

There are others, but that's sufficient for me to set out why I'm skeptical that switching to a purely chronological following feed is the silver bullet so many seem to think. (And, of course, to be pedantic, a chronological feed is an algorithm, just an exceptionally simple one.)

Let's do a quick thought experiment. Imagine you send a post out into the world on a social media platform. It's a criticism of the latest move by the Trump administration, where you remark that the move is clearly tied in with Trump's narcissism and cognitive decline.

The post does well. Reposts and likes start flooding in. You get replies and quote posts, some supporting and some negative. A fight breaks out in your mentions between your typical anti-Trump followers and a handful of MAGA types. Eventually, the engagement tapers off, but not before you pick up a bundle of new followers.

We can put a positive gloss on this. You wrote political commentary (that's valuable) that got attention (that's valuable) and provoked a bit of discourse (that's valuable) and netted you a larger audience (that's valuable). But we can also put a less positive gloss on it if we want. You wrote a quick post instead of deep analysis (you didn't elevate the discourse), you enraged some people (that's bad) and provoked arguments (creating toxic discourse), and your audience grew by adding followers who like toxic rage pages (leading to the risk of audience capture pushing you towards more toxic rage bait in the future).

If you take the positive side, and your post went out on a platform with a strictly chronological feed, you might frame it as getting some big accounts to repost you, which led to more eyeballs on your post, which led to valuable discussion. A win for the discourse and social media culture, the result of healthy chronological feeds. Everything that happened to get you lots of engagement is the result of what's good about sticking to an ordered timeline.

If you take the negative side, and your post went out on a platform with an algorithmic feed, it's that the algorithm knew your post would get a bunch of #resist "Right on! Take that, Drumpf!" shares alongside a lot of rage engagement from Trump fans, and so it pushed your post out to the wider world. Everything that happened shows exactly what's wrong with algorithmic feeds, in that they promote rage bait and created an algorithm feeding incentive for you to post it.

My argument, and why chronological feeds aren't a silver bullet, but instead just a way to make a complex problem seem simple, is that both stories can be correct because both styles of feed, in the end, are much less distinct than we'd like to think, and both incentivize similar kinds of bad behavior.

In an algorithmic feed, what you want to do is tune your posts to hit the algorithm's sweet spot. If that's engagement, then you write the kind of stuff that's engaging (rage bait is very engaging), and you post it frequently enough that you'll get enough algorithmic hits when the conditions are just right to take it into the vitality stratosphere. You'll then learn from the hits what precise combinations work and you'll double down on them. You'll fall, in other words, into audience capture, but with an audience of one: the algorithm.

The thing is, though, the algorithm is simply a way of mathematically modeling and predicting what the larger audience of actual people wants. The platforms want you to use them a lot. (If they're commercial, that is, but any sufficiently large platform has strong incentives to become commercial in one way or another.) They want you to use them a lot if they're selling ads because that's more ads seen, and so more ads sold. They want you to use them a lot if they're instead selling subscriptions, because you're more likely to maintain a subscription for a product you use frequently than one you don't. The reason the algorithm selects for shallow rage bait or other empty calories posts is because that's what most people want to like or share or click on.

With a chronological feed, then, assuming you, as a poster, still want engagement and follower growth (and most users do, unless they're entirely in it to lurk), you'll do the things that get you engagement and follower growth. And that means doing the things the audience wants. And to the extent the algorithm is correct in its guesses (and there's a lot of money riding on algorithms being correct in their guesses), it means doing things an algorithm would want you to do.

The way you go viral (i.e., get lots of engagement) in a chronological feed is to post the kind of stuff people will repost. Maybe the kind of stuff particular big accounts you know follow you are apt to repost. (Seeing as my most viral Bluesky posts have typically been kickstarted by

@kenwhite.bsky.social

reposting them, if I were engagement-maximizing, I'd write for Popehat.) Reposts both put your post in front of more people who will repost it and keep it alive, so new people keep seeing it, even if it's old enough that otherwise it'd be crowded out of their feed. This also means, if you want to consistently go viral and build an audience, with a chronological feed you'll post a lot. You can't count on a ranked feed to show your old posts, so you need to make sure you've always—always—got new posts. Yes, you keep your followers and so if you post slowly you still have the same number of potential viewers. But they're not all online at the same time, getting attention depends on those initial reposts, and growth begets growth, so if you take time off, you fall behind the posters who don't.

The upshot is that a chronological feed, assuming posters' desire for engagement is constant no matter the technical structure of the platform (and it probably is), will incentivize posting that is (1) reportable and (2) frequent. Put another way: flood the zone with rage bait. Which is what the folk theory tells us algorithms do.

This is all a (very long) way of saying that in most cases, what people blame on the algorithms is really just what does well on social media, regardless of the algorithm. Yes, you could tune an algorithm to select against it, but those tend to result in dull sites. Take, for example, Threads. What algorithm objectors object to isn't so much the algorithm itself as what it reveals about social media audiences. And so when they demand a chronological feed as the way to fix that, the unstated assumption is that, in the absence of an algorithm, social media users will have the same tastes they do. A chronological feed will show more of what they like and less of what they don't.

We can't escape the algorithm. Even if it's an algorithm of people, it's still to privilege what people like, even if it's the kind of stuff you want to see less of.

If you enjoy my writing, consider supporting me on Patreon. You'll get early access to all new episodes of my ReImagining Liberty podcast, as well. Learn more here.