The Dopamine Cycle: How TikTok's Recommendation Algorithm Shapes Minds?

I

n the digital age, social media platforms have transformed into tools of social engineering and informal educational spaces that shape the worldview of young people. Content promoted by algorithms has now transcended individual preferences and become a matter of social and national security. TikTok is a striking example of this. Launched in 2016 by the China-based company ByteDance, it surpassed 1 billion users worldwide within a few years and became the most popular platform among young people. The secret to its unique success lies in its special recommendation algorithm, which captivates users from the very first moment by offering an endless stream of videos. Unlike classic social networks, the platform offers an independent viral discovery mechanism, allowing videos from unknown content creators to reach millions.

However, perhaps the darker side of this algorithmic power lies in content selection and allegations of invisible censorship. Leaked communications from 2020 revealed that TikTok applied filters to ‘create a desirable vibe’, hiding users deemed unattractive, poor or disabled from the feed. Similarly, leaks revealed that videos with dilapidated or slum-like backgrounds were deemed ‘not cool enough’ and not shown to a wider audience. Interestingly, many news sites around the world that reported on these leaks were forced to remove their articles in 2021 and 2022. Without this information about the 2020 TikTok leaks, we might not have found these articles.

How does the TikTok algorithm work?

TikTok’s recommendation algorithm, which offers personalized video streams, is the most important factor that makes the app so popular and addictive. But how well is this algorithm developed in relation to neural networks? How does it trigger the neural system? Is it solely related to social psychology or is there an algorithm and content system compatible with the hormonal system? How does it create the cycle that causes a dopamine surge in the brain? Is there a new system that deciphers the neural system directly, replacing the film, television, and advertising industries shaped by archetypes? These are topics for other articles.

Returning to our research topic, the “For You” page on TikTok is designed to offer each user new videos tailored to their interests. When a new user joins the platform, the algorithm gathers interest signals by initially showing popular videos in different categories. Interactions, such as which videos the user watches and how much they like or share, quickly create a profile. If a user watches a video all the way through or rewatches it, the algorithm assumes they are interested in that content and recommends similar videos more frequently. Thus, TikTok’s recommendation algorithm creates a kind of “vulnerability map” of the user by analyzing every micro-interaction.

Although personalization enables users to swiftly discover content they may enjoy, it can also foster a hazardous cycle of dependency. Rather than being about digital platform loyalty, it is more accurate to say that this is an addiction involving the brain’s stimulation centers. Because the TikTok feed is designed for endless scrolling, users have difficulty controlling their content consumption. Each video is optimized to be short and attention-grabbing, constantly triggering the brain’s reward mechanism and releasing dopamine. Young people, especially children, have not yet fully developed impulse control, making it easier for them to get caught up in this rapid reward cycle and spend hours on the app. The platform constantly prompts users with the question, “What will the next video be?” This makes them lose track of time. Indeed, research shows that compared to other platforms, TikTok users lose control of their time more quickly and stay on the app much longer than planned.

This can be explained by TikTok’s psychological use of a “variable reward” principle. With each scroll, users experience uncertainty about what the next video will be. Although some mundane content appears occasionally, the frequent appearance of entertaining or interesting videos reinforces addictive behavior similar to pulling the lever on a slot machine in the hope of winning a big prize. The Washington, D.C., attorney general’s indictment also describes the TikTok algorithm as “intentionally addictive”, emphasizing that it keeps young people online for hours, which can lead to mental health issues such as depression, anxiety and attention deficit disorder.

Another concerning aspect of TikTok’s algorithm is its tendency to show users sensitive and harmful content. Various studies have revealed weaknesses in the platform’s content moderation. For instance, Global Witness’s 2025 experiment involved creating new accounts that appeared to belong to 13-year-olds. Even with “Restricted Mode” enabled, sexually explicit terms were suggested to these accounts in the search bar. After a few clicks, the platform itself suggested pornographic and inappropriate content. The suggestions were so inappropriate that researchers had to report videos featuring individuals who appeared to be under 16 to the relevant authorities for suspected child abuse. In short, not only does TikTok fail to fully implement the rules it claims to have introduced to protect young users but its algorithm also directs them toward objectionable content.

Persona experiment

To understand how the TikTok algorithm influences users, we conducted an experiment with fictional users who had different profiles and interests. We created 12 distinct personas that varied in age, gender and interests and opened a new TikTok account for each persona. These personas included children, teenagers, young adults and middle-aged individuals. Their interests ranged from humor and gaming to music, religious content, news, history and recipes. On the first day, each persona signaled their interests to the algorithm by searching for or watching a few related videos. For the next week, each persona passively watched content that appeared on their “For You” feed. On the third and seventh days, we recorded and compared the algorithm’s suggested content profiles for each persona. This allowed us to observe how the algorithm constructed user profiles and trapped them in specific content bubbles over the course of a week.

The findings at the end of the experiment were quite striking. In particular, the initial interests of children and adolescents quickly shifted towards inappropriate and potentially harmful content. For instance, a 13-year-old male persona initially watched gaming videos and innocent jokes but by the third day, his feed was dominated by crude humor and dangerous “challenge” videos. By the seventh day, he was being exposed to violent content, including street fights and the glorification of carrying knives and other sharp objects.

Similarly, a 15-year-old female persona initially watched popular dance trends and posts from high school students. Within a few days, however, the algorithm pulled her toward more sexually suggestive dance videos and erotically suggestive content. Even with “Restricted Mode” enabled, TikTok was showing sexually explicit search suggestions in the young account’s search section. After a week, the girl’s feed was filled with lewd jokes, exhibitionist dances and advertisements for adult dating apps. These examples clearly demonstrate how inadequate TikTok’s content filtering mechanisms are for children and teenagers.

Extreme content among young people

The algorithm was also seen pushing users toward increasingly extreme content, particularly among young adults and middle-aged individuals. An 18-year-old female user, who was initially interested in makeup tutorials and college entrance exam preparation videos, found that her feed was soon filled with videos promoting consumption and the luxurious lifestyles of popular influencers. By the seventh day, she was inundated with perfectionist beauty trends, cosmetic surgery advertisements and content glorifying wealthy lifestyles. This led her to question her own appearance and living standards.

A 30-year-old male user who initially consumed tech news and economic content was bombarded with conspiracy theory-like political videos by the third day. His feed was flooded with unverified and sensationalist claims about topics such as the pandemic or the global economic crisis. By the end of the week, this person was overwhelmed with ideologically polarizing and provocative content. In fact, an independent test by Guardian Australia reported that even a completely new TikTok account, created without any interaction, was quickly flooded with violent, far-right content and racist videos, which is parallel to our research.

One of our religiously inclined personas underwent a complete transformation. Initially, the 40-year-old conservative woman watched content related to religious discussions and family values. However, by the third day, TikTok began presenting her with increasingly nationalistic yet secular content. There was an increase in posts with racist and Turkish nationalist themes, and Islamic references were relegated to the background in the videos. After a week, this devout profile was confronted with harsh Turkish nationalist discourse that largely relegated religious identity to the background. These findings suggest that the platform prioritizes content that serves a specific agenda in Türkiye. Some news pages on social media have been accused in the press of being controlled by Zionist groups that promote “Turkishness without Islam.” Similarly, despite her religious sensibilities, this persona was exposed to anti-religious, nationalist content by the algorithm.