Meta's system cards show how AI powers content recommendations on Facebook and Instagram

AI

Meta Platforms Inc. said today it’s releasing a series of “system cards” that describe how the various artificial intelligence-powered content recommendation systems used by Facebook and Instagram work.

The move is designed to increase transparency about how AI surfaces content for individual users, based on their history of interactions, likes, interests and friends lists. In a blog post, Meta explained that Facebook and Instagram employ various AI systems to surface content that users will hopefully find useful and interesting. The goal is to ensure that people see more relevant and valuable content, as opposed to things that don’t interest them.

The company doesn’t rely on one, singular AI system to do this but rather a multitude of systems that work separately and sometimes in tandem with one another, to bring more interesting content into users’ news feeds. These systems, which work in real time, are powered by dozens of AI models that work to identify content and predict the likelihood that someone will be interested in it and want to interact with it.

“Our AI systems predict how valuable a piece of content might be to you, so we can show it to you sooner,” Nick Clegg, Meta’s president of global affairs, wrote in a separate blog post. “For example, sharing a post is often an indicator that you found that post to be interesting, so predicting that you will share a post is one factor our systems take into account. As you might imagine, no single prediction is a perfect gauge of whether a post is valuable to you. So we use a wide variety of predictions in combination to get as close as possible to the right content, including some based on behavior and some based on user feedback received through surveys.”

Meta states that it believes it’s important for users to be able to understand the technology that powers their experiences on Facebook and Instagram. As such, the 22 system cards released today, available in 22 languages, detail how each of its AI recommendation systems works, so users can better understand how AI helps to recommend the content they see. They provide deep insights into how people’s choices and behavior influence the type of content they see on their news feeds

For Facebook, there are 14 system cards that cover the Facebook Feed, Feed Recommendations, Feed Ranked Comments, Reels, Stories, Video, Notifications, Marketplace, Groups Feed, Individual Group Feed, Suggested Groups, Search, People You May Know and Pages You May Like. For Instagram, there are eight cards, covering Instagram Feed, Feed Recommendations, Stories, Explore, Reels Chaining, Search, Suggested Accounts and Notifications.

Facebook’s News Feed AI system card

“[The system cards] give information about how our AI systems rank content, some of the predictions each system makes to determine what content might be most relevant to you, as well as the controls you can use to help customize your experience,” Clegg explained.

Each AI system card contains four sections, including an overview of what the system does, a section that explains how it works, a section that describes how users can customize the types of recommendations created by the system, and finally a section that explains how it delivers content, including details on how that system informs the overall AI system used by Meta.

Meta explains that these AI systems rely on information such as the features of a specific post, and the user’s history of interacting with similar posts, in order to try and predict their level of interest in a certain piece of content. Altogether, Meta uses thousands of different signals across Facebook and Instagram to inform its AI systems. They include the number of likes, comments and views a post or video has received, how long or how often a user watches videos or views posts, and how the user interacts with a specific author or group.

In addition, the system cards also detail the controls that govern each AI system, so users can customize their experiences. When a user clicks to unfollow a certain author, temporarily hide content or “show less,” this action will inform the system to decrease the ranking score temporarily for similar types of content.

Clegg explained that the system cards are a key element of Meta’s commitment to openness, transparency and accountability. He noted that though many people are excited about the possibilities of AI, just as many have expressed concern about the risks of these systems.

“We believe the best way to respond to those concerns is openness,” he wrote. “… Companies should be more open about how their systems work and collaborate openly across industry, government and civil society to help ensure they are developed responsibly. That starts with giving you more insight into, and control over, the content you see.”

Meta said users can find all 22 system cards within its Transparency Center, and encourages people to use them to educate themselves and customize how AI powers their experiences on Facebook and Instagram.

Images: Meta Platforms