Filter Bubble

A filter bubble is the narrowed information environment produced when a recommendation system selects content based on a person's past behavior, gradually showing them more of what they already engage with and less of what they don't. Eli Pariser coined the term in 2011 to describe the effect as distinct from an ordinary echo chamber: an echo chamber is a social space someone chooses to enter, while a filter bubble is assembled for them by an algorithm they didn't consciously ask to curate their reality.

The mechanism is engagement optimization applied at scale: a feed ranks content by predicted click-through, watch-time, or interaction, and content that confirms an existing view (see Confirmation Bias) or triggers a strong reaction reliably outperforms content that challenges it. No one designs the system to produce ideological narrowing on purpose — it falls out of the incentive to hold attention, the same way water finds the lowest point. The effect compounds because the system also erases its own footprints; two people searching the identical term can get different results with no visible sign that filtering occurred, which is what makes the bubble hard to notice from inside it.

The empirical picture is more contested than the popular story: several large studies have found the average person's information diet is only modestly more homogeneous online than off, and that active choice (which outlets to follow, which communities to join) does at least as much of the narrowing as the algorithm does. The useful version of the concept isn't "algorithms make everyone extreme" — it's that personalization removes the friction that used to force people through a shared, unfiltered front page, and shifts more of the diet's shape onto choices, both the algorithm's and the person's own, that are individually invisible.