After Meta CEO Mark Zuckerberg said that they would more than double “recommended content” by the end 2023, Instagram revealed its algorithm to suggest new content.
On Friday, Instagram’s engineering team published a detailed overview on how the app’s artificial intelligence (AI) selects suggested posts that users see in their feed.
Instagram launched suggested posts in August 2020. The team stated that the goal was to duplicate human discovery of Instagram content using AI features in an effort to increase user engagement.
Instagram’s engineering team says that the app’s post recommendations fall into two categories: “Connected” and “Unconnected.”
In the connected system, Instagram’s AI finds and highlights recommended posts based on the user’s interests. This ranking system picks the most engaging posts according to factors such as relevance, quality and freshness.
In an unconnected system, sources are derived implicitly based on a user’s activity across Instagram and are then ranked based on similar factors.
The process is largely based on implicit signals from actions that users have directly taken in the app, such as following and liking posts. It can be extended to people and things that users like as a proxy of direct engagement. Some popular posts will also be highlighted depending on their overall engagement.
According to the engineering team the goal is to recreate the look and feel of Instagram profiles and posts when suggesting posts. This will make the experience more familiar.
“Scrolling through the End of Feed Recommendations should feel like scrolling down an extension of Instagram Home Feed,” the team says.
The team says that if users do not have enough recent engagement on Instagram for AI to suggest posts for them, the system evaluates accounts followed by that user and uses them as seeds for suggested content.
If the user is new to Instagram, the app simply shows them popular media items and adapts the parameters based on their initial responses.
Instagram’s engineering group says they try to make sure that the mix of media types (such as video and traditional post) is similar to suggested posts and Home. This message is in contradiction to Instagram’s decision to promote more Reels and video on its user feeds.
Instagram has faced wide-scale criticism from its users for its recent changes. In response to the growing backlash, Adam Mosseri from Instagram decided to reverse course and reduce temporarily the number of suggested posts so the team can improve its algorithm.
But Instagram users can expect to see the number of suggested posts steadily rise in the next year. During Meta’s second quarter earnings call last month, Zuckerberg revealed that, currently, about 15 percent of the content shown to a Facebook user on their feed is recommended by its algorithm — with the number being slightly higher on Instagram’s feed. However, according to Zuckerberg, this number should increase to more than 30 percent by the end of 2023.
“Social feeds are going from being driven primarily by the people and accounts you follow to increasingly also being driven by AI recommending content that you’ll find interesting from across Facebook and Instagram, even if you don’t follow those creators,” says Zuckerberg.