Research paper · 10 November 2026
Are Social Media Algorithms Controlling What We Think?
Recommendation algorithms decide what billions of people see every day. The research suggests they shape the information we meet far more clearly than they change what we believe.
By Kartikeya · 7 min read · 3 sources
Edited for publication and checked against the sources listed at the end.
An investigation into recommendation systems, human behaviour and the formation of public opinion.
Abstract
Social media has transformed the way people communicate, access information and understand the world. Platforms such as Instagram, TikTok, YouTube and Facebook use sophisticated recommendation algorithms to personalise the content displayed to each user. Although these systems make social media more relevant and engaging, they also raise important questions about their influence on human behaviour, personal opinions and public perception.
This research investigates how recommendation algorithms influence the information people encounter, whether they contribute to filter bubbles and echo chambers, and how much control individuals retain over their own opinions. Drawing on academic studies and survey research, this paper argues that algorithms significantly influence people’s digital environments, but their influence should not be confused with complete control over human thought.
1. Introduction
Imagine opening Instagram to watch a single video about technology. Within minutes, your feed is filled with technology-related videos, recommendations and discussions. Eventually, the platform begins suggesting content that closely matches your interests.
This experience is not accidental. Modern social media platforms use recommendation algorithms to determine which content appears on users’ screens. These systems analyse patterns in user activity and attempt to predict what individuals are likely to find interesting.
While personalised recommendations make platforms convenient, they introduce a significant question: if algorithms influence what people see every day, could they also influence what people believe?
Social media is no longer simply a platform for communication. It has become an important source of information, entertainment and public discussion. Understanding how its algorithms work is therefore essential to understanding the relationship between technology and society.
2. Understanding Social Media Algorithms
A social media algorithm is a computational system that helps determine which posts, videos and advertisements appear in a user’s feed.
Instead of displaying every available post chronologically, platforms generally rank and recommend content using different signals. These may include:
- Previous viewing and browsing activity.
- Likes, comments, shares and other interactions.
- Accounts and topics a user follows.
- The popularity and relevance of particular content.
- Predictions about whether a user will watch or interact with a post.
For example, if someone frequently watches videos about artificial intelligence, a recommendation system may begin displaying more content about programming, robotics and emerging technologies.
This creates a personalised digital environment. However, it also means that two people using the same platform can encounter very different versions of the internet.
The important issue is not simply that algorithms recommend content, but that they influence which information becomes visible and which information receives less attention.
3. The Filter Bubble: Are We Seeing Only What We Agree With?
One of the most widely discussed consequences of personalised recommendations is the filter bubble.
A filter bubble occurs when personalisation and individual content choices result in someone repeatedly encountering information that reflects their existing interests or viewpoints.
For example, consider two individuals with different opinions about a particular social issue. Their social media feeds may gradually become filled with content that supports their respective positions.
Over time, they may encounter fewer alternative perspectives, potentially making it more difficult to understand opposing arguments.
Closely related to this concept is the echo chamber. An echo chamber is an environment in which similar opinions are repeatedly reinforced through social interactions and shared information.
However, it is important not to assume that every personalised feed creates an echo chamber. People still choose whom to follow, what to watch and which information to trust.
The relationship between algorithms, personal preferences and social behaviour is considerably more complicated than it first appears.
4. What Does the Research Actually Say?
Research provides evidence that social media influences the information people encounter, although the extent of its influence on their beliefs remains debated.
A 2024 study by the Pew Research Center, How Americans Get News on TikTok, X, Facebook and Instagram, surveyed 10,287 American adults [1].
The research found that 90 per cent of TikTok users, 91 per cent of Facebook users, 92 per cent of X users and 82 per cent of Instagram users encountered at least one of the news-related content types examined in the survey [1].
Importantly, the research also found that two-thirds or more of the users of each platform saw opinions and funny posts about current events, even though most Facebook, Instagram and TikTok users said keeping up with the news was not a reason they used those platforms [1].
This demonstrates how social media can expose people to information and opinions they might not have actively searched for.
However, an important counterpoint comes from a 2025 study published in the Proceedings of the National Academy of Sciences (PNAS) [2].
The researchers conducted four experiments involving nearly 9,000 participants, on a purpose-built platform showing real YouTube videos and recommendations, to investigate whether politically biased recommendations could increase political polarisation [2].
Their findings showed that manipulating recommendations to create more politically one-sided content environments had limited detectable effects on participants’ political attitudes in the short term [2].
This finding challenges the assumption that recommendation algorithms automatically radicalise users or determine their political beliefs.
A further contribution comes from a September 2026 review published in Frontiers in Sociology. Researchers Sadiq C. P. and Rahul Shukla examined 60 studies published between 2020 and 2025 [3].
Their review found stronger evidence that recommendation systems influence the distribution and visibility of information than that they consistently produce ideological polarisation [3].
Together, these studies suggest that algorithms influence people’s information environments, but their effects on actual beliefs are more complex.
5. The Business Behind Recommendations
To understand why social media platforms use recommendation algorithms, it is necessary to examine their business models.
Many major social media companies generate revenue through advertising. Consequently, keeping users engaged with their platforms is commercially valuable.
Recommendation systems help achieve this by identifying content that users are likely to watch, interact with or share.
For instance, a user who regularly watches short comedy videos may receive similar recommendations because the system predicts that this content will maintain their interest.
However, this creates an important concern.
If platforms prioritise engagement, highly emotional, controversial or attention-grabbing content may receive significant visibility, depending on the platform’s ranking system.
This does not mean every platform deliberately promotes misleading information. It does, however, raise questions about whether commercial incentives always align with the public interest.
The challenge is finding a balance between creating engaging digital experiences and ensuring that users have access to reliable and diverse information.
6. Are We Losing Control of Our Own Opinions?
Despite the influence of recommendation systems, it would be inaccurate to suggest that algorithms completely control human thinking.
People interpret information differently depending on their experiences, education, personal values and social environments.
Two individuals can watch the same video and reach entirely different conclusions.
Furthermore, users retain some control over their digital experiences. They can follow different accounts, question information, explore alternative viewpoints and modify certain recommendation settings where platforms provide those options.
Nevertheless, the influence of algorithms should not be underestimated.
When people repeatedly encounter similar information, it can affect which topics they consider important and which viewpoints they become familiar with.
The issue is therefore not necessarily that algorithms tell people what to think, but that they can influence what people have the opportunity to think about.
7. Possible Solutions and the Future of Social Media
Reducing the negative effects of recommendation systems requires cooperation between technology companies, policymakers and users.
Some possible approaches include:
- Greater algorithmic transparency: Platforms should provide clearer explanations of why particular posts and recommendations appear in users’ feeds.
- More user control: Individuals should have meaningful options to adjust recommendations and access chronological or less personalised feeds where available.
- Digital literacy: Educational institutions should help students understand how recommendation systems work and how to evaluate information critically.
- Greater content diversity: Recommendation systems could provide more opportunities for users to encounter reliable information from different perspectives.
- Independent research: Researchers should receive appropriate access to platform information so that the effects of recommendation systems can be evaluated more effectively.
As artificial intelligence becomes increasingly integrated into digital platforms, these questions will become even more important.
8. Conclusion
Social media algorithms have fundamentally changed how people discover information, communicate and interact with the world.
They influence the visibility of content, personalise digital experiences and can contribute to environments in which similar opinions are repeatedly encountered. However, existing research does not establish that algorithms universally control people’s beliefs or automatically create political polarisation.
The evidence suggests that their strongest and most clearly demonstrated influence lies in shaping the information people encounter, rather than directly determining what they think.
Ultimately, the responsibility does not belong exclusively to technology companies or individual users. Both must recognise the influence of digital recommendation systems.
The most important question may therefore not be whether social media algorithms control our thoughts, but whether we understand how much influence they have over the information that shapes those thoughts.
The future of the internet will depend not only on how intelligently algorithms recommend information, but also on how responsibly that information is presented and consumed.
Research methodology: this paper uses secondary research, drawing on published academic studies, experimental research and survey findings. It does not present original experimental findings.
Sources
- How Americans Get News on TikTok, X, Facebook and Instagram — Shearer et al., Pew Research Center, 2024
- Short-term exposure to filter-bubble recommendation systems has limited polarization effects: Naturalistic experiments on YouTube — Liu et al., Proceedings of the National Academy of Sciences, 2025
- Consensus or conflict? Recommender systems as recursive exposure systems in the production of polarization — Sadiq and Shukla, Frontiers in Sociology, 2026