Published on

How TikTok's Algorithm Figures You Out | WSJ

How TikTok's Algorithm Figures You Out | WSJ

Introduction

It seems like TikTok knows us better than we know ourselves. The uncanny accuracy with which TikTok's For You page curates content has led many users to wonder just how the app does it. They even express concerns that TikTok might be listening to their conversations. To demystify how TikTok gets to know its users so intimately, the Wall Street Journal undertook an extensive investigation. They created over a hundred automated TikTok accounts, or bots, to observe how the app's highly secretive algorithm functions.

The Experiment

To understand TikTok's algorithm, the Wall Street Journal set up these bots to mimic real users with various interests ranging from extreme sports to depression. These bots never explicitly expressed their interests but interacted with content by re-watching or pausing on videos related to their assigned topics.

How TikTok Deciphers Your Interests

TikTok’s engagement starts broadly by showing users a wide range of popular videos, many vetted by moderators. It then narrows down based on user interactions. Thus, the length of time a user lingers on a piece of content becomes the primary signal that informs TikTok's algorithm. Even though TikTok does not read your texts or listen to your conversations, the app collects significant cues from your content interaction, allowing it to decipher hidden user interests.

Results from the Bots

Several key findings emerged from this experiment:

  1. TikTok can fully understand users' interests in less than two hours, sometimes even within 40 minutes.
  2. As the algorithm gets to know you, the videos you see become more specific and niche, often with fewer views and less oversight.
  3. Bots interested in mental health, relationships, and anxiety issues quickly received a flood of content related to these themes, often diving deeper into a singular topic.
  4. Even bots with diverse interests could find themselves in narrow content bubbles if their interactions signaled stronger interests in specific topics.

Experts Weigh In

Guillaume Chaslot, a former Google engineer who worked on YouTube's algorithm, reviewed the Wall Street Journal's findings. According to Chaslot, algorithms on platforms like TikTok and YouTube are optimized to catch user vulnerabilities, leading them to content that maximizes screen time, which doesn't necessarily mean the content is beneficial for the user.

The Potential Pitfalls

While TikTok includes features like a "not interested" button to guide content recommendations, the algorithm's primary design is to maximize engagement, sometimes to the detriment of the user. Bots that were steered into content related to conspiracy theories, sexual content, or mental health issues could easily become stuck in these less-moderated content loops, potentially encountering harmful content.

Conclusion

Our daily interactions with algorithms shape both us and the technology we use. Understanding how these algorithms function is increasingly crucial to protect user well-being and ensure fair practices. Through this study, the Wall Street Journal highlights an essential dialogue about transparency and ethics in algorithmic recommendations.

Keywords

  • TikTok Algorithm
  • Content Recommendations
  • User Interaction
  • Mental Health
  • Engagement
  • Social Media Algorithms
  • Content Moderation

FAQ

Q: How does TikTok figure out your interests? A: TikTok primarily uses the length of time you interact with a video, including pauses and re-watches, to gauge your interests.

Q: Is TikTok listening to my conversations to serve content? A: No, TikTok does not eavesdrop or read your text messages. It relies heavily on user interactions with the app.

Q: How quickly can TikTok understand my interests? A: The algorithm can understand your interests in less than two hours, sometimes in as little as 40 minutes.

Q: What happens if I keep watching videos about a specific topic? A: If you consistently interact with content on a specific topic, TikTok will show you more content related to that theme, often pushing you deeper into niche rabbit holes.

Q: How does TikTok manage harmful content? A: TikTok uses a combination of computer analysis and human moderation to review videos. Users can also report inappropriate content, but harmful content can still slip through the cracks, especially in highly personalized content niches.

Read More