How Algorithms Are Quietly Shaping the Way We Think

Have you ever opened TikTok, Instagram, YouTube, or another social platform “for five minutes” and somehow ended up watching completely different videos half an hour later?

It feels like we are choosing what to watch. And technically, we are. But there is something happening quietly in the background: Algorithms are choosing which options we see in the first place.

That does not mean an algorithm is sitting somewhere inside your phone, telling you what to believe. It is much subtler than that.

Every click, like, search, share, pause, skip, and viewing habit can become a signal. Recommendation systems use these signals to predict what you might want to see next. YouTube, for example, says its recommendations use signals including watch and search history, likes, dislikes, subscriptions, “Not interested” feedback, and even satisfaction surveys. 

In other words, your feed gradually becomes a reflection of your behavior.

Watch five videos about a subject, and you may receive more. Watch ten, and suddenly your feed can begin to feel as if that subject is everywhere.

This can be incredibly useful.

Algorithms help us discover musicians we might love, educational videos we would never have found, new creators, breaking news, recipes, hobbies, and communities. Without recommendation systems, the internet would be an enormous library with no librarian.

But there is a catch.

The same system that helps us discover things can also narrow what we discover.

Imagine you watch several videos supporting one particular opinion. The system learns that you are interested. It may then show you more related content—not necessarily because it wants to change your beliefs, but because that content has become more relevant to your predicted interests.

Eventually, your feed may begin to look like the entire world agrees with you.

But the real world is rarely that simple.

This is where the idea of an “echo chamber” becomes important. When people repeatedly encounter similar opinions, their existing beliefs can become stronger while opposing perspectives become less visible.

Research is increasingly examining this relationship. A 2026 study using computational modeling found that recommendation-style connections can increase social fragmentation and contribute to echo-chamber effects under certain conditions. That does not prove that every algorithm automatically makes people polarized, but it does show why the design of recommendation systems matters.

Politics makes the issue even more interesting.

In June 2024, Pew Research Center reported that 66% of U.S. news consumers on X and 61% of TikTok news consumers said the platforms influence which news stories they see at least somewhat. 

And in 2025, researchers published an audit of TikTok recommendations during the 2024 U.S. presidential election. Their experiment involved 323 accounts and roughly 394,000 videos viewed between April 30 and November 11, 2024. They reported differences in the political content recommended to accounts seeded with different partisan material. The study is one piece of evidence—not a final verdict on TikTok—but it illustrates why recommendation systems are increasingly being studied as part of the information environment around elections. 

There is another important detail: Algorithms are not necessarily trying to make you believe something.

Usually, their immediate job is closer to predicting what you will watch, click, or find satisfying. YouTube explicitly describes its system as trying to help viewers find content they are likely to enjoy and return to. 

That distinction matters.

The question is not simply, “Is the algorithm manipulating us?”

A better question might be:

Because attention has consequences.

If dramatic content gets more attention, creators have an incentive to make dramatic content. If outrage generates discussion, outrage can become profitable. If short videos keep people scrolling, platforms have a reason to make the next video effortless to watch.

Over time, the system and the user can influence each other.

We teach the algorithm what catches our attention, and then the algorithm gives us more of what catches our attention.

It becomes a loop.

Fortunately, users are not completely powerless.

Platforms provide ways to influence recommendations. YouTube, for instance, allows users to mark videos “Not interested,” remove watch history, manage search history, and adjust recommendation signals.

Regulators are also paying more attention. In the European Union, the Digital Services Act has introduced transparency requirements around recommender systems, including obligations concerning their main parameters and user access to options that are not based on profiling in relevant cases. Harmonized transparency rules began applying on July 1, 2025, with the first standardized reports published in February 2026.

So, are algorithms quietly deciding what we think?

We still have our own judgment, experiences, values, conversations, and ability to question what we see.

But algorithms increasingly influence what reaches our attention—and what reaches our attention can influence what we know, discuss, fear, admire, and eventually believe.

Maybe the most important form of digital independence is therefore not deleting every social media app.

It is learning to pause.

To ask, “Why am I seeing this?”

To search beyond the first recommendation.

To deliberately read something we disagree with.

And occasionally, to close the feed altogether.

Because the algorithm may know what keeps us watching.