If your social feeds feel eerily personal, it’s because they are. Every pause, like and swipe teaches the system what to show you next. Over time, it doesn’t just reflect your interests. It reshapes them.
Researchers are beginning to map just how quickly this happens. A University of Washington team analyzed more than nine million TikTok recommendations. Within the first thousand videos, the app had already locked onto users’ preferences and started serving content designed to maximize their time on the platform. Over a four-month period, average daily use climbed from half an hour to nearly an hour.
“The platform designs are not neutral,” lead researcher Franziska Roesner said. “They influence how long you watch and what you watch, and what you’re getting angry or concerned about.”
That design is intentional. Studies of YouTube show users are often pulled toward increasingly sensational material. Mozilla’s crowdsourced research found people routinely regretted what the algorithm recommended to them, especially when it pushed misleading or extreme videos. TikTok studies have documented similar patterns, with teens in particular reporting addictive use and lower self-esteem after long exposure. The point isn’t that these platforms directly cause harm. It’s that their recommendation systems consistently reward content that keeps us hooked, whether or not it’s healthy.
Kyle Chayka, author of Filterworld: How Algorithms Flattened Culture, says the effect goes beyond screen time.
“Algorithmic recommendations are addictive because they are always subtly confirming your own cultural, political and social biases,” he writes. “Warping your surroundings into a mirror image of yourself while doing the same for everyone else.”
That mirror is comforting, but it erodes culture. What used to feel like discovery now feels like déjà vu. The same interiors, the same worship songs, the same viral reels.
“In passively consuming what I was interested in, had I given up my agency to figure out what was truly meaningful to me?” Chayka asks.
The research on political polarization tells a similar story. A major Facebook study found its News Feed slightly reduced exposure to opposing views, but users themselves filtered out even more by choosing who to friend and what to click. Other studies of YouTube suggest the platform can reinforce radicalization, but not always in the dramatic ways critics assume. In most cases, the algorithm amplifies paths we’re already on.
Chayka argues that amplification is what hollows out culture. “All kinds of cultural experiences have been reduced to the homogenous category of digital content and made to obey the law of engagement,” he writes. What thrives are things that are easy to replicate and quick to consume. What fades are the messy, slow or difficult experiences that once gave culture depth.
Regulators are paying attention. In the U.K., Ofcom has announced new requirements aimed at platforms that push harmful material to children. “We will not tolerate designs that funnel minors toward damaging content,” CEO Melanie Dawes said earlier this year. The new Online Safety Act will force companies to rework the recommendation systems themselves, not just tack on labels.
Some researchers are trying to prove these systems can be re-tuned. A Stanford team built a prototype feed that downranked anti-democratic content. Users who tried it showed lower partisan hostility, with no measurable drop in engagement. “If we can make a dent in this very important value, maybe we can learn how to use social media rankings to affect other values we care about,” said co-author Michelle Lam.
Still, public trust is thin. Pew surveys show most Americans expect AI to make news worse, not better. Another study found skepticism about using AI to handle decisions that feel personal, like relationships or creative work. People may scroll endlessly, but they also sense the cost of letting the machine steer.
Which is why Chayka’s suggestion of an “algorithmic cleanse” resonates. His advice isn’t to go off-grid, but to notice how much the feed has shaped your life and then reclaim some agency. Subscribe to a magazine. Buy a record in person. Watch a movie because a friend recommended it, not because it landed on your homepage.
“You may not use social media,” he warns, “but it’s using you.”
It’s easy to dismiss that as alarmist until you count the ways your mood shifts with your feed. How many conversations start with, “I saw a video about…” How many evenings dissolve into clips that feel personal but leave no trace. The algorithm didn’t steal your attention. It bargained for it, and most of us accepted the deal without thinking.
The question is whether we’re willing to renegotiate.
