In the midst of its first professional development event for creators, Instagram has dropped what appears to be an unusually transparent look at its content-surfacing algorithms.
“It’s hard to trust what you don’t understand,” platform head Adam Mosseri said in an official blog post. “There are a lot of misconceptions out there, and we recognize that we can do more to help people understand what we do.”
Chief among those misconceptions is the idea that Instagram’s content suggestion sections and features–such as individual users’ feeds and the platform’s Explore tab–are all steered by a single, omnipotent algorithm, Mosseri says.
Subscribe for daily Tubefilter Top Stories
Instead, Instagram uses “a variety of algorithms, classifiers, and processes, each with its own purpose,” to surface content, he explains.
Some of these systems were implemented as part of Instagram’s generally unpopular decision to change users’ feeds so they no longer displayed every post from people they followed in chronological order.
Mosseri says Instagram made this change because internal data showed that as the site’s user base and amount of user-generated content grew, “it became impossible for most people to see everything, let alone all the posts they cared about.” By 2016, users were not seeing 70% of posts in their chronological-order feeds.
It’s hard for people to trust what they don’t understand, which is why we wanted to shed more light on how Instagram works, and why you see what you see. We often get asked about “the algorithm,” so we wanted to break things down a bit more.
— Adam Mosseri 😷 (@mosseri) June 8, 2021
So, Instagram now organizes users’ feeds based on four key classifiers. They are, in order of importance: information about the content itself, such as amount of engagement, what time it was posted, and whether it’s a photo or video post; information about the person who posted it, including “signals” that show how interesting they are to other users; user activity, including what topics they might be interested in; and last up, the interaction history between two users (so, if a user comments regularly on another person’s posts, that person’s posts could be more likely to appear on their feed).
Instagram employs these classifiers to predict what kind of feed content and Stories each user could be interested in, and then surfaces that content for them. It uses the same set of classifiers, but in a slightly different order of importance, to serve content on Explore.
Instagram uses similar classifiers to rank Reels–but prioritizes entertainment
Mosseri also delved into how Instagram ranks content from its TikTok competitor Reels.
“Much like Explore, the majority of what you see is from accounts you don’t follow,” he says. “With Reels, though, we’re specifically focused on what might entertain you.”
Again, Instagram uses the same four classifiers, but with Reels, users’ activity is the No. 1 influence on what content they’re recommended.
“We survey people and ask whether they find a particular reel entertaining or funny, and learn from the feedback to get better at working out what will entertain people, with an eye towards smaller creators,” Mosseri says. “The most important predictions we make are how likely you are to watch a reel all the way through, like it, say it was entertaining or funny, and go to the audio page.”
Mosseri additionally, briefly addressed another contentious topic: shadowbanning. People have long theorized that Instagram (and Facebook, and YouTube, and Twitter, and TikTok) secretly prevent some users’ content from reaching others, while not formally suspending or terminating their accounts.
But, Mosseri claims, this isn’t a thing on his platform. To people who suspect they were shadowbanned because their number of likes or comments went down, he says Instagram can’t promise steady engagement, because “[t]he truth is most of your followers won’t see what you share, because most look at less than half of their Feed.”
That being said, Mosseri reveals Instagram is currently developing more thorough in-app notifications for users whose content violates its Community Guidelines, with the goal of better helping people see when the company has taken action on their posts.