On social media, high school students and parents can follow the same movement while perceiving very different realities. By selecting content based on each person’s habits, algorithms reinforce distinct, sometimes opposing, news feeds and make it harder to understand the other person’s point of view.
The same movement, two news feeds
When a protest movement spreads through high schools, social media becomes one of the places where it is discussed and commented on. Videos of school blockades, images of police interventions and students’ statements circulate quickly. But they do not appear in the same way on every screen: each user sees a selection shaped by their previous browsing and interactions.
Outside Jacques-Decour High School, in northern Paris, Lina, 17, is in her final year. Since the protests began, her TikTok feed has mainly been showing her videos related to the blockades and high school students denouncing police violence. She knows these posts are often biased and are intended to defend a particular point of view.
On Facebook, her parents see more images of damage and of people portrayed as rioters. Worried, they share them with their daughter to warn her about the blockades. The two generations are therefore talking about the same event, but are drawing on content that shows different aspects of it.
The differences also appear in how the movement’s prominent figures are portrayed. Lina is more likely to receive videos that are favorable to Elwan, a leading figure in the protests. Her parents more often come across clips of him losing his temper or making a mistake during a TV interview. So it is not only the facts that differ from one feed to another: the image of the people at the center of them can also change.
How algorithms choose what we see
Platforms aim to offer posts likely to hold users’ attention. To make their recommendations, they take many signals into account: videos watched to the end, posts shared, accounts followed, comments posted and even content users linger over. Gradually, these choices build a profile of users’ interests and shape what they are shown next.
This mechanism can create a repetition effect. When someone often watches videos presenting a particular interpretation of an event, the platform may show them more videos like them. At the same time, opposing viewpoints or information less in keeping with their habits may become less visible in their feed.
Personalization does not mean that all users are trapped in a completely sealed bubble. They may encounter different opinions, particularly when a post becomes very popular or someone close to them shares it. But the daily selection tends to give more prominence to posts that already match their online behavior.
This logic can also affect discussions about intelligence and thinking habits. A reflection on the effects of social media and how the people around us influence our ideas is offered in this article about social media, intelligence and the influence of those around us.
The perception gap between generations
David Chavalarias, a researcher at the CNRS and author of Toxic Data, describes a perception gap between different groups. Members of one side may encounter the other side’s representatives through particularly caricatured clips or moments of controversy. What circulates in this way does not always give a balanced picture of their arguments or motivations.
Among high school students, posts may highlight the reasons for the protests, students’ accounts and images of rallies. For parents, the most visible content may instead emphasize risks, incidents or the consequences for safety and education. Each selection makes its own interpretation of the movement more familiar and can give the impression that it is the most representative one.
Warnings from family members may then be interpreted as a refusal to understand the demands. Conversely, parents may feel that their children are downplaying the risks because they do not see the same images as they do. Social media does not create disagreements on its own, but it can deepen misunderstandings by giving each person different examples to support their perception.
When content becomes evidence
A video shared in a feed can end up serving as the central piece of evidence in a discussion. Yet a short clip does not necessarily provide the full context: it may begin after an important event, end before what follows, or pair images with a highly biased commentary. When users regularly see similar content, these fragments can add up to a coherent story in their eyes, even if other information changes how it should be interpreted.
Fast-paced formats also encourage immediate reactions. A dramatic scene or a provocative statement is more likely to attract attention than a long, nuanced explanation. More emotional posts can therefore take up a significant place in news feeds, while contextual details are shared or viewed less often.
This circulation has a particular consequence for conversations between parents and teenagers. Forwarding a video to warn or persuade someone does not guarantee that the recipient will interpret it in the same way. Depending on their experience, beliefs and the content they are used to seeing, they may view it as a legitimate warning, an exaggeration or an example of a partisan point of view.
Can we escape our algorithmic bubble?
Taking back some control over your feed
Thibault Philippette, a professor of information and communication sciences at UCLouvain in Belgium, encourages users not to let recommendations alone determine their information journey. The more passively someone views the content suggested to them, the more similar signals platforms have to suggest similar posts.
A first step, then, is to actively seek out information. It is possible to consult several news outlets, check where a video came from, look for the context of a clip and compare accounts of the same event. This approach will not instantly change a feed, but it helps users avoid mistaking a platform’s recommendations for a complete picture of current events.
Users can also question their own habits: which accounts provide them with information? Which points of view are rarely represented? Is a post being shared because it offers new details, or because it confirms an opinion they already hold? These questions help distinguish genuine source diversity from the mere repetition of similar content.
For families, a conversation can begin by comparing news feeds. Parents and teenagers can explain what they saw, where a post came from and why it seemed important to them. The goal is not to decide immediately who is right, but to recognize that each person may have received a different and incomplete selection.
Why a few interactions are not always enough
A parent may be tempted to watch or like several videos about high school students in order to change their recommendations. However, suggestions are based on accumulated habits, not just a few recent interactions. An account that has long viewed other categories of content will not necessarily turn quickly into a feed like that of a young protester.
Changing recommendations can take time, and there is no guarantee that two people will see the same posts. Algorithms continue to interpret each account’s behavior based on its history and reactions. The bubble may shift gradually, but people’s digital journeys often remain different.
Changing platforms to change the rules
David Chavalarias argues for a more radical option: leaving platforms whose systems encourage polarization and turning to networks designed according to different principles. Mastodon, Bluesky and Eurosky are among the services mentioned in this context. However, they have far fewer users than major platforms such as TikTok or Instagram.
These networks do not promise to make disagreements disappear or to guarantee perfectly balanced information. Their value lies instead in design choices and recommendation systems that can limit certain attention-grabbing mechanisms. Their smaller audiences also mean users may not find the same volume of videos, accounts or discussions there as on the dominant platforms.
The limited presence of these services in people’s daily lives may contribute to their lower visibility. Very popular platforms benefit from network effects: people go there in part because their friends, family and the public figures they follow are already there. By comparison, less addictive or less widely adopted networks can seem less engaging, even when they offer another way to follow current events and conversations.
Diverse sources in everyday life
Escaping a bubble does not depend solely on a setting or switching networks. It also means varying the sources you consult and paying attention to the mechanisms that make some content more visible than others. A video that appears frequently in a feed is not necessarily representative of an entire movement, just as a selection of alarming images does not sum up all the realities surrounding a blockade.
In discussions between generations, comparing the content each person receives can reveal the blind spots in each feed. Parents can discover the accounts that help students make sense of the protests; high school students can learn about the images and concerns that worry their families. This comparison does not eliminate disagreements, but it helps explain how algorithmic bubbles contribute to the formation of parallel narratives.
