How TikTok and Instagram Use Artificial Intelligence: Does the Algorithm Decide What We See?
Discover how AI and algorithms personalize social media and the risks of feed manipulation. Who controls what we see on social platforms?
Algorithms decide what we see on social media
We scroll with a finger and, as if by magic, perfect videos appear for us. But how does TikTok know we like funny cats or cooking tutorials? And why does Instagram keep showing us posts about a topic we only glanced at?
The answer lies in artificial intelligence. The algorithms that govern social media feeds are not simple lines of code: they are complex systems that learn from our behaviors to offer us personalized content. A convenience, certainly. But also a potential tool of manipulation that deserves to be thoroughly understood.
How recommendation systems work
Every time we like a post, spend more time on it, or watch a video to the end, we are giving valuable information to a recommendation system. This system memorizes our behavior, compares it with that of millions of other users, and tries to understand what we might want to see next.
This is how the famous "personalized bubble" is built: a continuous stream of tailor-made content, optimized to keep us glued to the screen. TikTok's algorithm, for example, considers over 1,000 different signals for each user: from time spent on videos to scrolling speed, from social interactions to app usage times.
Instagram uses a similar but more layered approach, combining data from the main feed with that from Stories and Reels. The result is a content ecosystem that seems to know us better than we know ourselves.
The artificial intelligence behind personalization
The technology that makes all this possible is based on increasingly sophisticated machine learning algorithms. These systems use deep neural networks capable of identifying complex patterns in our digital behaviors.
TikTok, in particular, has developed an algorithm called "For You Page" that combines collaborative filtering (suggestions based on similar users) and content-based filtering (analysis of content characteristics). The artificial intelligence analyzes not only what we watch, but also how we watch it: micro-movements of the finger, pauses, even the angle of the phone.
Instagram has integrated AI in an even more pervasive way. Beyond feed content, the algorithm also influences search results, suggested content in Explore, and even the order of Stories. As we explored in our article on AI and social media: algorithms that guide us, these systems are redefining the way we consume information online.
Concrete Examples of Algorithmic Manipulation
The problem is not so much the efficiency of these systems, but their unintended side effects. We risk being exposed only to opinions similar to our own, reinforcing beliefs without any counterpoint. In the worst cases, we can be led towards extreme, conspiratorial, or manipulative content, not because the algorithm "wants" to do it, but because it has learned that this type of content keeps us engaged longer.
An investigation by the Center for Humane Technology highlighted how social platforms inadvertently incentivize the spread of polarizing content. TikTok, for example, has been criticized for how its systems can quickly push users towards radical videos or conspiracy theories if the user shows even minimal initial interest.
Facebook (now Meta) has admitted that its algorithms tend to favor content that generates "engagement," even when this means amplifying anger or outrage. In 2021, Frances Haugen's testimony before the U.S. Congress revealed how the company was aware of these effects as early as 2018.
Privacy and Data Collection: The Price of Personalization
All of this happens thanks to a sophisticated combination of machine learning, predictive analytics, and massive collection of personal data. And this is where the privacy issue comes into play. The data we surrender to social media – even just by interacting with a piece of content – is processed to reconstruct our tastes, vulnerabilities, and emotional tendencies.
TikTok collects over 380 different types of data on its users, according to an analysis by cybersecurity researcher Felix Krause. Instagram is no different: through its in-app browser, it can track every click, every cursor movement, even texts typed but not sent.
The stated goal is to keep us active and present for as long as possible. But at what price? As we explored in our article on focus and attention in the digital age, this hyper-connectivity is having profound effects on our ability to concentrate.
Algorithmic Bias and Digital Discrimination
What's concerning is not just profiling. So-called algorithmic biases also play a crucial role. Algorithms are not neutral: they learn from human data, which is often imperfect. If a certain type of content has been rewarded in the past, it will continue to be, reinforcing existing trends and penalizing diversity of perspective.
This is how dynamics are built that favor certain groups and marginalize others. Instagram, for example, has been accused of penalizing content created by people of color or belonging to minorities. TikTok has admitted to using policies that limited the visibility of creators with disabilities or those deemed "non-conventional" to avoid bullying, effectively creating a form of discriminatory censorship.
A study by MIT Technology Review demonstrated that TikTok's algorithm tends to show different content to users of different ethnicities, even when they have similar interests, perpetuating social divisions through personalization.
Key Points to Remember
- Our digital behaviors fuel increasingly sophisticated algorithms that create personalized content bubbles
- Personalization can lead to radicalization when it prioritizes content that generates strong emotional reactions
- The data collected goes far beyond likes and shares, including micro-behaviors and psychological patterns
- Algorithmic biases reflect and amplify human prejudices, creating systemic discrimination in the digital world
Frequently Asked Questions
How can I reduce the influence of algorithms on my feeds? Actively diversify your interactions, follow accounts with different opinions, and regularly use the "Not Interested" function when appropriate.
Do social media platforms really know that much about me? Yes, they collect hundreds of different data points, often cross-referencing information from multiple sources to create detailed profiles of your interests and behaviors.
Are there alternatives to traditional social media? More transparent platforms like Mastodon or BeReal are emerging, which use simpler algorithms or chronological feeds, but they still have limited adoption.
How do I know if I'm in a filter bubble? Check if you rarely see opinions that contradict your beliefs or if your social feeds are very homogeneous in the themes and viewpoints presented.
Towards Greater Digital Awareness
Artificial intelligence in social media has two faces. On one hand, it allows us to discover new content, connect with those who share similar interests, and experience smoother digital interactions. On the other hand, it can become a distorting lens, showing us only a part of reality—the one that keeps us engaged.
To tackle this complexity, awareness is needed. Digital literacy is required to help people recognize the underlying mechanisms, to question why we see certain content and not others. As we discussed in our in-depth analysis on fake news and information warfare, the capacity for critical thinking becomes increasingly essential.
Only in this way can we transition from passive users to critical digital citizens. The future of social media depends not only on technology, but on how we decide to use and regulate it. And on how much we want to understand what lies behind every scroll.