The algorithmic feed will get regulated

Every day, a handful of ranking models decide what billions of people see. They aren’t built to inform us. They’re built to keep us scrolling.

I think it makes us subject to manipulation at the scale of whole societies, partly by accident and partly on purpose. The accidental part needs no villain. The model learns what holds our attention, and outrage, fear and envy hold it best. Facebook knew. A 2018 internal presentation said it plainly: “Our algorithms exploit the human brain’s attraction to divisiveness.”

It works because of how our brains are built, in two ways.

First, we react to danger before anything else. For most of history, missing a threat could kill you, and missing good news rarely did. So a post that frightens or angers us grabs our attention faster than one that informs us.

Second, we want to belong. In The Status Game, Will Storr describes how we are driven to join groups and to earn status inside them. Outrage at the other side is a cheap way to show which group you belong to, and your group rewards it with likes and shares. A study of more than half a million tweets found that each moral or emotional word in a tweet raised its retweets by about 20%, mostly within the author’s own political camp.

You may recognise the result in the swing of the pendulum. A few years ago, identity politics on the left dominated the conversation online. Today, it is populism on the right. Each end feeds on outrage at the other, and outrage is what the feed rewards. I don’t think the feed created either side. But it amplifies whichever side makes us most afraid or angry at the moment, and that makes each swing faster and harder.

It also distorts how we see each other. The loudest and angriest voices on each side are served up as if they were everyone. So we come to believe the other side is more extreme than it is. Americans on both sides think about twice as many of their opponents hold extreme views as actually do. The divide grows out of a picture that was never true.

Then there is the intentional part. The model works for anyone who learns to game it. Before the 2016 US election, a foreign troll farm ran fake American accounts on Facebook, Instagram and Twitter. In Houston, it organised two opposing rallies outside the same Islamic centre on the same day, one for each side. In Europe, fake copies of Le Monde and Der Spiegel have spread stories about the war in Ukraine. And in December 2024, Romania’s constitutional court annulled a presidential election after declassified intelligence pointed to a coordinated TikTok campaign behind a little-known candidate who had won the first round.

None of these campaigns created a divide. They found one and made it louder. Interference can only amplify what is already there, and the feed is the amplifier. Whoever owns a platform holds the volume knob, and can turn it whenever they like.

Right now, the debate is about children. Australia banned social media for under-16s in December 2025. Denmark wants an age limit of 15. The European Commission has preliminarily found that TikTok’s addictive design breaks EU law.

Good. The damage to young people’s mental health is real, and it is massive. But it is dwarfed by what the same feeds do to all of us: what we believe, how we vote, and who we think the enemy is. An age limit protects a 14-year-old. It does nothing for the feed their parents scroll every night.

The obvious answer is to ban social media altogether. I think that’s wrong. Social media is the closest thing we have to a public square. It’s where anyone can say what they think and be heard, and where we keep in touch with people we would otherwise lose. A ban would be a ban on free speech and on being connected.

The problem isn’t that people can post. It’s the machine that decides who sees what. So I think regulation will separate speech from reach, in three steps.

First, the default feed becomes the one you chose: the people you follow, newest first. The EU already requires the largest platforms to offer a feed that isn’t based on profiling. Today you have to go looking for it, and some argue it is hidden on purpose. Next, it becomes the default, and the engagement-ranked feed becomes something you switch on.

Second, you get to pick the algorithm. The platform hosts the posts and the connections. The ranking can come from someone else. Bluesky already works this way with custom feeds. The ranking could even come from your own agent, working for you instead of for advertisers. It’s one more job for the box in the cupboard.

Third, whoever ranks at scale answers for what they amplify. A newspaper editor decides what a few hundred thousand people read, and answers for it under media law. The feed is the most powerful editor in history, and it answers to no one.

The current situation can’t continue, and people will realise it, the way they realised that lead in petrol was poisoning everyone. We didn’t ban cars. We took the lead out of the petrol. We don’t need to ban social media. We need to take the engagement ranking out of the feed.