Article by: Olivia Carling, Open Dialogues International Foundation
Social media is often presented as a neutral tool that simply shows users the content they want to see. In reality, platforms such as TikTok, Instagram, YouTube, Facebook, and X (Twitter) are powered by algorithms designed to maximize engagement. Their primary goal is to keep users scrolling, clicking, sharing, and watching for as long as possible. While this business model is highly profitable, growing research suggests that it can also amplify misinformation, reinforce harmful stereotypes, and deepen existing social inequalities.
Algorithms are not inherently racist, sexist, or classist. However, they learn from human behavior and historical data. When those data reflect existing prejudices, the algorithms can reproduce and scale those biases to millions of people. Combined with engagement-driven recommendation systems, this creates a digital environment where misinformation and divisive content often outperform factual and balanced information.
One of the most significant problems is the way algorithms reward emotionally charged content. Studies examining social media recommendation systems have found that content provoking anger, outrage, fear, or hostility tends to generate higher engagement rates than neutral information. Because algorithms interpret engagement as a sign of quality or relevance, they often amplify controversial posts regardless of whether they are accurate. Research published in EPJ Data Science found evidence that X's recommendation system amplified low-credibility content more effectively than higher-credibility information. The study also found increased visibility for highly toxic and politically biased posts, demonstrating how recommendation systems can unintentionally contribute to the spread of misinformation.
This dynamic has real-world consequences. During major political events, public health crises, and social movements, false or misleading information can spread faster than verified reporting. Researchers studying misinformation in the United Kingdom found that approximately one in ten social media users regularly engages in sharing exaggerated or false news stories. While this may appear to be a minority, the scale of modern social media means that even a small percentage of users can significantly influence public opinion.
Algorithms also create what researchers call "filter bubbles" or "echo chambers." By analyzing a user's viewing history, likes, comments, and shares, platforms continuously recommend content that aligns with existing beliefs. While this personalization improves user engagement, it can limit exposure to opposing viewpoints and reinforce pre-existing opinions. Over time, users may become increasingly convinced that their perspective represents the majority view because they rarely encounter alternative perspectives. This process contributes to political polarization and weakens democratic dialogue by encouraging people to consume information that confirms rather than challenges their beliefs.
The impact of these systems extends beyond politics. Increasingly, researchers and educators are expressing concern about how recommendation algorithms shape young people's views on gender, race, and class.
One widely discussed example is the rise of misogynistic content targeting teenage boys. Influencers associated with the so-called "manosphere" have gained enormous visibility through algorithmic recommendations on platforms like TikTok, YouTube, and Instagram. Figures such as Andrew Tate became some of the most searched and recommended personalities online despite repeated criticism for promoting sexist ideas and degrading attitudes toward women.
Andrew Tate on his podcast: Tatecast
A recent Common Sense Media study found that 69% of boys aged 11 to 17 reported regular exposure to online content promoting harmful ideas about masculinity. These messages often portray men as dominant, emotionally detached, and entitled while presenting women as inferior or responsible for men's problems. Importantly, many boys reported being shown this content even when they had not actively searched for it. Recommendation systems identified engagement patterns among similar users and pushed the content into their feeds.
Teachers across the United Kingdom and Ireland have reported growing concerns about the influence of misogynistic influencers on students. Some educators describe hearing boys repeat phrases and ideas popularized by online personalities, often disguised as jokes or irony. While many young users may initially engage with this content for entertainment, repeated exposure can normalize harmful attitudes. Over time, jokes about women being inferior, manipulative, or responsible for social problems can become accepted as common sense rather than recognized as prejudice.
The same mechanisms that amplify misogyny can also reinforce racism. Algorithms frequently prioritize engagement over social responsibility, meaning content that provokes outrage or controversy may receive increased visibility. Throughout social media history, creators have built large audiences by making racist remarks, promoting stereotypes, or using racial controversy as entertainment. Because such content often generates strong reactions from both supporters and critics, it can receive greater engagement and therefore greater algorithmic amplification.
At the same time, research suggests that creators from marginalized communities often face additional barriers. Studies on shadowbanning and algorithmic visibility have found that women, LGBTQ+ creators, plus-size creators, and other underrepresented groups frequently experience reduced visibility, content suppression, or inconsistent moderation. Automated systems sometimes incorrectly flag cultural content, activism, or discussions about discrimination as inappropriate. As a result, the very groups most affected by bias may find it harder to have their voices heard online.
Classism is another less discussed but equally important issue. Social media algorithms frequently reward content associated with wealth, luxury lifestyles, and consumerism because such content attracts attention and aspirational engagement. Influencers displaying expensive vacations, designer clothing, luxury cars, and lavish homes often receive significant algorithmic promotion. Meanwhile, the realities of poverty, economic insecurity, and working-class experiences tend to receive less visibility. This creates a distorted picture of society in which wealth appears normal and universally attainable, contributing to unrealistic expectations and feelings of inadequacy among viewers.
Above, Charlie D’Amelio, ‘Mr Beast’, and Emma Chamberlain, some of the richest influencers in the world
For young users, these effects can be particularly powerful. Adolescents are still forming their identities, beliefs, and understanding of the world. When recommendation systems repeatedly expose them to misogynistic influencers, racist stereotypes, misinformation, or unrealistic lifestyles, those messages can gradually shape how they think about social issues and other people. What begins as entertainment can become a framework for understanding gender roles, race relations, economic success, and political identity.
Importantly, researchers emphasize that algorithms do not create these social problems from nothing. Instead, they often reinforce existing societal tensions, prejudices, and inequalities. The problem is that social media allows these patterns to spread faster, reach larger audiences, and become embedded in everyday digital experiences.
As social media becomes increasingly algorithm-driven, the challenge facing technology companies is clear. Platforms must decide whether their systems will continue prioritizing engagement above all else or whether they will begin designing algorithms that value reliability, diversity of perspectives, user well-being, and social responsibility. The future of digital media will not only determine what people see online, it will help shape how future generations understand the world itself.
Bibliography
Cecere, G., Jean, C., Le Guel, F. and Manant, M. (2024). Artificial Intelligence and Algorithmic bias? Field Tests on Social Network with Teens. Technological Forecasting and Social Change, [online] 201, p.123204. doi:https://doi.org/10.1016/j.techfore.2023.123204.
Draper, D. (2023). The Pros and Cons of Social Media Algorithms. [online] Available at: https://bipartisanpolicy.org/wp-content/uploads/2023/10/BPC_Tech-Algorithm-Tradeoffs_R01.pdf.
Jonker, A. and Rogers, J. (2024). What Is Algorithmic bias? [online] IBM. Available at: https://www.ibm.com/think/topics/algorithmic-bias.
Metzler, H. and Garcia, D. (2023). Social Drivers and Algorithmic Mechanisms on Digital Media. Perspectives on Psychological Science, [online] 19(5), pp.735–748. doi:https://doi.org/10.1177/17456916231185057.
Moore, M. (2025). Meet the Richest Influencers in the World – Including a 12-year-old Child. [online] HELLO! Available at: https://www.hellomagazine.com/celebrities/825736/richest-influencers-world/.
Putri, S.D.G., Purnomo, E.P. and Khairunissa, T. (2024). Echo Chambers and Algorithmic Bias: The Homogenization of Online Culture in a Smart Society. SHS Web of Conferences, [online] 202(1), p.05001. doi:https://doi.org/10.1051/shsconf/202420205001.
Singer, A. (2022). The Negative Effects of Social Media Algorithms. [online] THE PENNSYLVANIA STATE UNIVERSITY SCHREYER HONORS COLLEGE. Available at: https://honors.libraries.psu.edu/files/final_submissions/8388.
