Filter Bubble Lab is an educational web app that lets you operate the post feed of a fictional social network called Flock to see, firsthand, how recommended content gradually becomes biased. No sign-up is required, it is free to use, and nothing you do here is sent to an external server. The point of this activity is to let you feel, through your own actions, something that is hard to grasp from a written explanation alone: the sense that your own choices are what shape the feed in front of you.
The eight topics you can choose from are pets, lifestyle habits, reading, games, health and diet, AI and work, a fictional city's redevelopment, and space and science. Some topics, such as pets or lifestyle habits, split clearly into two sides, like being pro-cat or pro-dog, or an early riser or a night owl, while others, such as health or science, do not center on taking a side. Whichever topics you choose, the flow of the activity and what you see on the reveal screen stay the same.
The feed's ranking is a simplified model of how real social ranking systems behave. The three topics you chose at the start receive a higher initial score, and liking, sharing, or following a post raises that topic's score further, while marking a post "Not Interested" lowers it. For posts that take a side, such as being pro-cat or pro-dog, liking or sharing them also gradually strengthens your learned leaning toward that side. On top of this, posts that already have many likes, and posts written with anxious or absolute-sounding language, receive an extra boost regardless of whether their content is accurate, reproducing how engagement-driven ranking can reward attention-grabbing posts. A small amount of randomness is mixed in each time so the feed does not become completely fixed. Until you have been shown a post from every topic, the feed intentionally spreads across all topics before ranking by learned score takes over.
Once you have taken enough actions (roughly 15 actions, or 40 displayed posts, whichever comes first), you unlock a reveal screen that lets you look back at what happened to your feed. It walks through five perspectives. First, a bar chart shows how strongly the system learned to favor or avoid each topic based on your actions. Second, it charts a diversity score from 0 to 100, calculated from how mixed the topics were among the ten most recent posts, so you can see whether the second half of your session grew more repetitive than the first. Third, it shows posts that existed on Flock at the same time but were never shown to you at all. Fourth, it lines up the feed that a version of you with the opposite learned preferences would have seen, illustrating how the same network can present a very different world depending on your actions. Finally, it reports how many low-reliability posts you were shown and how many of those you liked or shared, as a reminder to pause before reacting to attention-grabbing content.
After the reveal, five lesson cards walk through the underlying ideas: what a filter bubble is, how it differs from an echo chamber, confirmation bias as a human tendency, what recommendation algorithms are actually optimizing for, and five concrete habits you can start today, consulting multiple sources, diversifying who you follow, remembering that marking something Not Interested is also a learning signal, checking sources before sharing, and occasionally searching for viewpoints yourself instead of waiting for a recommendation. The lessons are followed by a five-question, four-choice quiz tied directly to what you just experienced, covering what triggered your feed's bias, the warning signs of an unreliable post, the difference between a filter bubble and an echo chamber, what signals the activity's algorithm actually relied on, and the most effective response once you notice bias in your feed. Depending on your score, you receive one of three titles, Bubble Observer Beginner, Bubble Watcher, or Algorithm Trainer, and can then replay the same experience with a more deliberate eye.
Filter bubble and echo chamber are often confused, but they refer to different mechanisms. A filter bubble refers to the system side: how a search engine or social platform selects what to show based on your browsing history and reactions. An echo chamber refers to the human side: how people with similar opinions repeat the same views to one another until those views come to feel like a majority position. Either can occur on its own, but system-driven selection and social reinforcement often overlap, deepening the same bias from two directions at once. This activity's feed and reveal screen mainly reproduce the system side, the filter bubble. It is worth remembering that on a real social network, echo-chamber dynamics among the people you follow are usually happening alongside it.
The low-reliability posts in the feed were deliberately written to share certain traits: strong claims with no cited source, urgent language such as "share this now," extreme numbers with no evidence behind them, wording designed to provoke anxiety or rivalry, and hints of rumor or conspiracy. On the reveal screen you can check exactly which of these signs appeared in each post, so you can turn the list into a quick checklist for posts you come across on real social media.
Everything runs with mouse or touch input only, requires no account, and sends nothing to an external server, so it works directly on a school device or a personal phone. Topic selection takes only a moment, and the feed stage moves you forward once you reach 15 actions or 40 displayed posts, so it does not drag on indefinitely. Including the reveal screen, the five lessons, and the five-question quiz, the whole activity is sized to take roughly 15 to 20 minutes, though this varies by person. For a single class period, students can work through the activity individually and then discuss their diversity scores and their engagement with low-reliability posts as a group. Useful discussion prompts include why a student's diversity score fell or stayed steady, how it felt to see how many low-reliability posts they reacted to, and how their learned interest profile compared with a classmate's.
There are a few simple checks you can try after finishing this activity. You could look back at your own likes and follows on a social platform you actually use, and notice whether your feed already leans toward one direction. When a news item catches your attention, you could look for the original source instead of judging it from the headline alone. You could also try following just one account with a different perspective from the ones you usually see. Each of these puts the habits covered in this activity, diversifying who you follow, checking sources before sharing, and searching things out yourself, directly into practice on a real social network.