Recommendation Systems in Social Media
The rating
Our assessment of how this product aligns with each of The Institute’s eight AI Principles. Full detail in the Evaluation section below.
AI Principles
Recommendation systems, often referred to as recommenders, are algorithmic systems designed to suggest relevant content or products to users. They are at the heart of our online experiences, shaping what we see and interact with on platforms like social media, news sites, and streaming services.
This review focuses on recommenders in social media platforms, which determine the content that Instagram shows you from your connections and other accounts, for example, or the videos you see in TikTok's "For You" feed. For social media platforms like these, recommenders are essential to both user engagement and platform revenue.
Recommendation systems wield strong influence that can affect users' behavior and beliefs. To maximize user engagement, they guide what people see by presenting content that feels personally relevant and engaging. As a result, users may return frequently if the experience feels rewarding and familiar. Prolonged engagement with personalized content can limit exposure to diverse perspectives and viewpoints.
Recommendation systems have already caused real world harms, including online radicalization, echo chambers, and harmful recommendations like eating disorder content. Research and documented cases show that these systems can unintentionally steer users toward narrow, repetitive, or even harmful content, especially when many platforms are not upfront with how content is prioritized.
Kids and teens are heavy users of social media , but these platforms aren’t designed with their development in mind. While research on social media’s impact is mixed, young users face unique risks. Kids and teens are often treated as adults by recommendation systems to increase their engagement, which can alter their opinions, preferences, and behavior in unexpected and unforeseeable ways.
Recommendation systems suggest relevant content or products to users, often to maximize user engagement. They sort, group, and prioritize information on platforms like social media, news sites, and streaming services and are trained to predict how a user will react to different types of content. Recommendation systems can help users find new and interesting content, help users connect with new people around shared interests, and help users grow a following online.
While they can be helpful, it’s not clear how they track users and recommend information which can create risks and harms. There is no common framework or widely agreed-upon process for deciding which signals are tracked and prioritized, how they are measured, or how trade-offs should be evaluated across platforms. Similarly, there are few standards or rules for how recommendation systems can be used in social media.
What Recommendation Systems in Social Media get right
Recommender systems can be helpful in many ways, such as:
By suggesting new content that is truly aligned with your interests, these algorithms can help you find new and interesting people, topics, and information.
Recommendations can help you connect with new people around shared interests, or learn new things.
Who doesn't love finding a video of a dog's very special relationship with his toys that you may never have seen without a social media platform recommending it to you?
Recommenders can supercharge the content you create, which can help you grow a following or influence public discourse in meaningful ways. This can be a great equalizer.
Where they fall short
Players have no idea when a machine is going to hit it big, so they keep playing. Recommenders serve relevant content, and the urge to constantly check them out and scroll the feed is the social media equivalent of pouring dollar after dollar into a slot machine. In fact, researchers studying the effect that social media has on the brain have shown that social media exploits "the same neural circuitry" as "gambling and recreational drugs to keep consumers using their products as much as possible." This is in large part because of the recommendation systems driving the content shown to users.
There are documented cases of online radicalization facilitated by recommender systems, and of users being exposed to violent or harmful recommendations (for example, recommenders promoting "corpse bride" diets to girls, advocating 300 calories a day). They can also be far too easily exploited by bad actors in ways that have created serious harm.
These occur when users are only exposed to content that aligns with their existing beliefs and preferences. This can limit exposure to diverse perspectives and viewpoints, hindering meaningful discourse and fostering polarization. And because they are so powerful, they are central determiners of information, which gives them enormous power over societal opinions. Despite this, many platforms are not transparent about what signals they're using to surface recommended content and how they prioritize those signals.
Recommenders have the potential to influence users' well-being through the content they recommend. Studies have shown conflicting evidence regarding the impact of social media on well-being. Some studies suggest positive effects such as increased life satisfaction and social trust, while others show decreased mental health and increased addiction tendencies. Common Sense's research has shown that it's often both.
This means that children and teens are, as one researcher notes, "increasingly treated as adults in a context where recommendation systems, by virtue of their aim to increase user engagement, have the capacity to fundamentally alter their opinions, preferences, and behavior in ways neither they, their guardians, or system designers can foresee."
More risk assessments
