The algorithm decides now

Two people sit at the same kitchen table, on the same morning, scrolling the same social feed. One closes the phone convinced the world is on fire — furious, certain, braced for a fight. The other closes it calm, having seen recipes, a science clip, and a friend’s holiday photos. They didn’t choose different worlds. A machine chose for each of them, story by story, and neither was told it was happening. That machine is the new gatekeeper. It does not think like an editor. It thinks like a slot machine.

The gatekeeper isn’t a person anymore

In lesson 2 you met the gatekeeper: an editor deciding what reaches the front page. For most of history that gatekeeper was human, slow, and at least nameable. On a social feed, the gatekeeper is a ranking algorithm — a piece of software that, every time you open the app, sorts thousands of possible posts and decides which handful you actually see, and in what order.

Here is the thing to hold onto. The feed is not a list of everything, newest first. It is a ranked list, and the ranking is a choice. The software scores every candidate post and shows you the top ones. So the only question that matters is: what is it scoring them for?

It optimises for engagement, not truth

The answer is engagement — the umbrella word for anything you do with a post. Time spent looking at it. A tap. A like. A share. A comment. A reply to a comment. The algorithm has one job baked in by its owners: keep you doing things, because doing things keeps you on the app, and time on the app is what gets sold to advertisers. That’s the attention business from lesson 8, now wearing a smarter face.

So the feed promotes whatever scores high on engagement. And here is the trap: what’s engaging is not what’s true, and not what’s important. Those are different qualities, and the machine can only measure one of them. It cannot tell whether a claim is accurate. It can easily count whether you stopped scrolling.

What reliably makes people stop, tap, and argue? Things that are emotional, surprising, or identity-confirming. Above all, things that make people angry — outrage is the strongest fuel there is, because nothing drives comments like something to argue against. A calm, careful, true post sits there earning almost nothing. A furious, sloppy, half-false one lights up the scoreboard. Same shelf, very different scores.

A worked example: ranking the feed

Picture six posts waiting to be shown to you. Each has an engagement score — roughly, how many people stop and react to it:

• A measured explainer on a new policy — score 12
• A local charity hits its fundraising goal — score 9
• A furious post claiming a group “is destroying everything” — score 95
• A practical guide to fixing a common problem — score 14
• A shocking (and unverified) rumour about a public figure — score 78
• A nuanced piece weighing two sides of a hard issue — score 11

A human editor weighing importance might lead with the policy explainer. The algorithm doesn’t weigh importance. It sorts by the only number it has. So the furious post (95) goes to the very top of your feed. The unverified rumour (78) comes next. The careful, important posts — 12, 14, 11 — sink to the bottom, where almost nobody scrolls. They effectively vanish.

Now the loop tightens. You see the angry post first, and — being human — you stop, you read, maybe you tap or comment to disagree. The machine just learned something: this person engages with outrage about that group. So the next batch it builds for you leans further that way. More outrage, fewer explainers. Do that for a week, a month, a year, and your feed is no longer a window on the world. It’s a mirror of your own strongest reactions, fed back to you and amplified.

Two worlds, one feed

This is why the two people at the kitchen table saw different worlds. The algorithm learns what each person engages with and serves more of it. Your feed is personalised — built for you alone — which sounds like a gift and works like a cage. You mostly meet things you already agree with, said louder. People who think differently get sorted into a different feed entirely. This is the filter bubble: a personalised slice of reality where the disagreeing voices quietly disappear, not because anyone banned them, but because they didn’t score well with you.

And because the system rewards engagement over accuracy, false and sensational content has a built-in advantage. A lie crafted to enrage can out-compete a truth that asks you to think. Not because the lie is more true — because it is more engaging, and engagement is the only race being run.

The whole, ranked for you

Step back. No human editor sat down and decided you should see the furious post first. A scoreboard did, optimising for a number that has nothing to do with whether the post is true or whether it matters. The choice is still being made — lesson 1’s filter is still there — but now it’s automatic, invisible, and tuned to you specifically, by a system whose goal is your attention, not your understanding.

You are not above this. Knowing how the ranking works doesn’t lift you out of the feed — you still scroll it, still feel the pull of the outrageous post, still get sorted. But you can hold one quiet fact while you read: no one chose this for being true. It was chosen for keeping me here. That single doubt is the gap between being ranked and noticing that you’re being ranked. The next lesson names the slants this leaves behind — the kinds of bias you’ll learn to spot.