How posts get into recommendations and “shadow ban”: the head of Instagram explained the work of the social network’s algorithms

 

The system determines the interest rating of posts according to users and authors, on the basis of them makes predictions and forms the feed.

Instagram CEO Adam Mosseri published a video explaining how the algorithms work.

Adam Mosseri, chief executive officer of Instagram Inc., speaks during the Samsung Electronics Co. Unpacked launch event in San Francisco, California, U.S. on Wednesday, Feb. 20, 2019. Samsung debuted its most extensive new lineup of smartphones, taking on Apple Inc. amid a slowing market with new low-end and premium models, 3-D cameras, an in-screen fingerprint scanner and faster 5G connectivity. Photographer: David Paul Morris/Bloomberg via Getty Images

“In 2010, Instagram was a timeline of square photos. In 2016, we noticed that people were missing 70% of the content they were subscribed to. So we decided to change the system so that users would only see the most interesting stuff. Instagram uses multiple algorithms,” is how Mosseri began to describe the system.

Ranking in the feed and Stories

The system determines which post to show a user based on different factors:

  • The system evaluates the posts of all the accounts the user is subscribed to.
  • The rating of interest for the user is based on information about the post and the author: the published photo or video, how many likes, how close the user knows the author, what the author has published before, which posts the user reacts to most often, how often the user reacts to the author’s content, how long the post holds the user’s attention, how often the user leaves the author’s posts for the profile.
  • The publication time of the post is not the most important criterion, so the feed is approximately chronological.

Ranking in recommendations

The system identifies groups of topics that may be of interest to users based on several criteria:

  • Which posts a user typically likes and shares.
  • If you are subscribed to a popular culinary account, the system compares how many people who are also subscribed to it like another culinary account, and if many, may recommend it to you.
  • Also, the system analyzes information about the post and its author, the user’s interaction history with the author’s account, and the user’s activity with different types of content.
  • The system predicts with what probability the user will like, share or save the post – these are the main criteria for getting into recommendations.
  • Ranking in Reels also works.
Read also:   Instagram tests sticker for resharing in stories

 

“Shadow Ban.”

Often users say “shadow ban” when they don’t achieve the reach, likes and reactions they expect. Mosseri says Instagram can’t guarantee authors the same response to all posts because people’s interests change.
Mosseri admits that Instagram’s system can make mistakes in predicting users’ interests, which can make the reach and reactions less.

There are also specific topics that Instagram doesn’t welcome. For example, hateful posts, violent scenes.

How to influence your Instagram feed

  • The system will show more content from those you’ve added to your list of close friends.
  • You can hide posts that aren’t interesting.
  • If you mark a post “Not Interested,” the system will show less of them.

Leave a Reply

Your email address will not be published. Required fields are marked *