Do filter bubbles and echo chambers really exist?
The evidence is more nuanced than the popular account, and the honest answer is that both effects are real but smaller and more complicated than the standard narrative claims.
The original argument: personalised ranking shows you what you already agree with, so you progressively stop encountering opposing views, and society fragments into isolated informational worlds.
What research has actually found:
People on social media encounter more diverse political content than people who avoid it, not less. Several large studies have found that social platforms expose users to a wider range of sources than offline news habits or direct visits to a favoured outlet, because weak ties in a social graph carry different material.
Self-selection matters more than the algorithm. Who you choose to follow, and what you click, shapes your feed more than the ranking system does. Large-scale experiments that switched users to chronological feeds found smaller effects on polarisation than expected.
A minority is genuinely siloed, and it matters. Highly politically engaged users cluster far more tightly, and heavy consumers of partisan content do inhabit something like an echo chamber. This is a real phenomenon in a small, highly active and disproportionately influential group.
Exposure is not persuasion. One important finding is that encountering opposing views can increase entrenchment rather than reduce it, so the assumption that more diverse feeds would depolarise is not well supported.
Where the concern remains well-founded: amplification of emotionally provocative content, the ease of finding and reinforcing fringe communities, and recommendation systems surfacing progressively more extreme material in some documented cases.
Practical steps if you want a broader view: follow deliberately across perspectives, use lists or chronological feeds where available, and read primary sources rather than commentary — but do not expect a feed change alone to change how you think.