Social media, algorithms and filter bubbles

Most people today no longer encounter news on a newspaper page, but in a feed, an endless stream put together by an algorithm. To check sources, you have to understand how this stream comes about.

The attention economy

Social platforms are usually free because we are the product, more precisely: our attention. Platforms earn from advertising, and advertising earns from time spent. The goal of the recommendation algorithms is therefore engagement: clicks, likes, comments, watch time. Not truth, not balance, not your well-being.

This has consequences. Content that triggers strong emotions (outrage, fear, enthusiasm) generates more engagement and is therefore spread more widely. A peer-reviewed study (Milli et al., PNAS Nexus 2025) showed that engagement-based ranking on X (Twitter) specifically amplifies emotionally charged, hostile content directed against the other political side, and this even though users did not actually prefer such content themselves and felt worse afterwards (study).

Remember

The algorithm does not optimise for what is true, but for what won't let us go. Outrage holds us well, so outrage is rewarded.

Influencers and the blurring of advertising

On social media the line between information, entertainment and advertising blurs. Influencers feel like familiar acquaintances, but they are often business partners of brands. Paid content must be labelled as advertising, but is not always. The old question helps: Cui bono: who benefits from this post? And: would this person say the same thing if they weren't being paid for it?

Filter bubble and echo chamber: what the research really shows

Two terms shape the debate:

• Filter bubble (coined by Eli Pariser, 2011): personalisation algorithms show everyone „their own information universe“ and screen out anything that contradicts. Crucially: by definition the filter bubble arises through algorithmic filtering, not through one's own choice.
• Echo chamber: an environment in which mostly like-minded voices reverberate. It can arise through algorithms, but also through one's own choice (we follow those we agree with).

Here is an important twist that calls for critical thinking: the popular filter-bubble thesis is empirically surprisingly weakly supported. A literature review by the renowned Reuters Institute (Oxford) reaches the following findings:

• True echo chambers are rare. In the United Kingdom only about 2 to 5 % of people live in a politically one-sided news echo chamber; most have a fairly diverse media diet.
• Algorithms tend to lead to more diversity, not less, through incidental encounters with sources one would never have chosen oneself. That is „the opposite of what the filter-bubble thesis claims“.

Source: Ross Arguedas et al., Echo chambers, filter bubbles and polarisation: a literature review, Reuters Institute (link).

Caution

A catchy theory is not true simply because it is repeated everywhere. The filter bubble is a good example: the image is immediately plausible, but the data hardly support it. A object lesson for critical thinking.

What remains of the algorithm problem?

Does this mean algorithms are harmless? No. The concern merely shifts to the right understanding: the robustly documented problem is not isolation in the bubble, but the amplification of outrage, hostility and emotional escalation. We don't get too little opposing opinion; we get the opposing opinion presented to us in its most annoying, most polarising form.

Practical consequences

• Don't mistake the feed for the world. What you see is an algorithmic selection, not a representative cross-section.
• Break out deliberately. Head straight for reputable sources on purpose, instead of just waiting for the feed. Actively seek out different sources.
• Pause when stirred up. Strong emotion in the feed is often a sign that a piece of content has been optimised for reach, so it's time for SIFT.
• Sharing is spreading. Every click on „share“ is a vote for the algorithm. Check first, then share.