How the Algorithm Decides What You See — And What That Costs You

Social media companies want to maximise engagement, which drives advertising revenue. But "engagement" is an abstract concept. To train a computer to maximise it, engineers must give the machine specific, measurable targets—known as proxies.

Instead of telling the algorithm "show people high-quality content," they tell it to predict:

• Will this user stop scrolling to look at this image?
• Will they click "like"?
• Will they leave a comment?
• Will they share it?

The algorithm then tests millions of posts against these proxies. Over time, it learns exactly what combination of imagery, emotion, and formatting triggers those measurable actions.

The cost of optimising for reaction

The problem with measuring success through clicks and comments is that human attention is most reliably captured by strong emotions—specifically outrage, anxiety, and novelty.

A thoughtful, nuanced essay might be deeply rewarding to read, but it rarely triggers the immediate, impulsive "share" or angry comment that a polarizing hot take does. Because the algorithm rewards what gets a reaction, it systematically filters out nuance in favour of extremes.

Over time, this curation distorts our view of the world. It makes extreme opinions seem mainstream. It makes conflict feel inescapable. It replaces quiet, slow reflection with a constant, exhausting demand for an immediate reaction.

How to train your feed (and yourself)

You cannot turn off the algorithm, but you can change how you interact with it. Because the system learns from your behaviour, you have to stop giving it the signals it craves.

1. Stop hate-reading

Every time you slow down to read a post that makes you angry, the algorithm learns that anger keeps your attention. If you want a calmer feed, you have to scroll past the outrage bait without slowing down.

2. Use the "Not Interested" button

Most platforms bury a "mute" or "not interested" option in a menu. Use it aggressively. It sends a much stronger negative signal than simply scrolling past.

3. Seek out chronological spaces

Move some of your reading to places where algorithms have less power. RSS feeds, email newsletters, and direct group chats allow you to see what you subscribed to, in the order it was published, without a machine interposing itself.

Attention is a zero-sum game

Your time and attention are finite resources. Every hour spent reacting to a manufactured controversy in an algorithmic feed is an hour you did not spend reading a book, learning a skill, or talking to a friend.

The algorithm is not evil; it is just doing exactly what it was built to do. The question is whether what it wants is aligned with what you want. Usually, it isn't.

Backed by clinical sleep lab trials, here is why banking sleep on the weekend fails to reverse weekday cognitive deficits, disrupts your circadian clock, and impairs metabolic health.

19 October 2026 · 5 min read