Echo Chambers and Algorithm Patterns

The familiar scroll

If you open your phone to your preferred social media app and scroll, a video will appear that perfectly matches exactly what you were thinking about earlier. The next one will too. And the next. The comments will agree with you. The people you follow — they think the same way. An idea will begin to feel like an obvious truth.

But how did you get here? Did you choose the information, or did the algorithm choose it for you? This is called an “algorithm echo chamber,” and it becomes surprisingly difficult to tell where our own choices end and the algorithm's begins.

A feed that learns

Your feed learns from you. The social media platforms use recommendation systems that take signals from our behavior on their platform — what we like, share, search for and scroll past — and use this to predict what we will likely engage with next. This creates a personalized experience, but it is also a powerful business tool.

The tech industry seems to view humans as a finite resource. To maximize profit, they use AI-powered algorithms to find the content you gravitate towards, personalize it, and feed it back to you in two harmful ways.

The PRIME engine

Firstly, the algorithms are engineered to exploit the brain's dopamine pathways, and purposefully amplify PRIME content (prestige, related, moral, emotional; all highly engaging content) in order to hold attention and keep people scrolling longer by keeping their emotional temperatures hot. At extremes, they may promote “ragebait” or “moral outrage” content. Interestingly, a May 2026 study in Sage Journals shows this phenomenon of modern echo chambers in this framework:

1. 01Ignition (Stage 1): “incendiary” or provocative content that goes viral.
2. 02Amplification (Stage 2): comments and reshares accelerate circulation.
3. 03Stabilization (Stage 3): institutional normalization into the mainstream public and in offline communities.

Next, the algorithm uses data from the content you consume and like, and the creators you follow, and creates an optimized personal “echo chamber” or “filter bubble.” I was intrigued by Story Penncock's experiment where she created two separate Twitter accounts and followed left-leaning and right-leaning channels on each, which resulted in two completely opposite echo chambers, showing how the algorithm feeds you reinforced, identical and increasingly radical viewpoints based on your behavior, all the while filtering out opposing content. This can have disastrous outcomes, seen in the example of Robert Stanford, whose consumption of more and more conspiracy theories ultimately led him to join the January 6th attack. One may see this as propaganda that no longer requires government takeover, but instead simply creating highly emotional and divisive content to easily manipulate and further polarize the masses.

Two identical phones

Imagine doing the same experiment with two teens, who begin with identical phones, apps, and connections. One spends a week watching videos on the dangers of artificial intelligence, and one spends a week watching videos about the benefits of artificial intelligence. Their feeds will begin to look completely different, yet neither person has necessarily been lied to. This makes the situation more complicated, because each person has received true information, just a selective portion of it.

If everyone in your feed agrees with you, it is easy to believe everyone outside your feed is uninformed or irrational. Personally, I find it terrifying when thought of in this way. Other examples of the harmful effects of echo chambers are the manosphere and the anti-healthcare / anti-vaxxer communities.

Are we really trapped?

The phrase “echo chamber” makes it sound as if algorithms are putting everyone in a digital prison, but this is not the case. A Reuters Institute literature review found that politically partisan online echo chambers are uncommon, and only a small minority of people rely exclusively on partisan sources. It also found that algorithmic selection does not consistently make people's news diets less diverse; in some studies, algorithmic recommendations actually exposed users to somewhat more diverse sources. So the algorithm must not be the entire problem.

We choose our own information environments — who to follow, whose videos to watch until the end, which titles and headlines to click on, and what information to block. Algorithms are responding to our choices. So what if we control the algorithm? Platforms like Threads, Instagram and TikTok are increasingly giving users tools to customize their recommendation systems. This allows users to either tighten the echo chamber, which is appealing because it is more comfortable to not have beliefs questioned or challenged — or, used deliberately, to crack one open.

Open question for the next instalment: what are people, companies and services actually doing to combat this?