In the blogpost, Mosseri also explained why the posts on users’ news feeds are not in a chronological order.
Instagram algorithm: After being criticised for long over its allegedly biased algorithm, Instagram has now decided to try and explain to some extent how the platform decides what the users will see. In a blog post, Instagram head Adam Mosseri said that there were a lot of misconceptions among people regarding the social media platform and he wanted to explain its system better. The first issue that Mosseri addressed was that of “algorithm”, clarifying that Instagram did not have one algorithm, but instead used a variety of processes, algorithms and classifiers, with each fulfilling a purpose of its own.
In the blogpost, Mosseri also explained why the posts on users’ news feeds are not in a chronological order. Back in 2010 when it was first launched, the platform used to push out images in a chronological order, but with more and more people and pages becoming active on the platform, the system stopped working for the platform, Mosseri said, citing that by 2016, users had been missing out on 70% of the posts on their Feed, including missing half of the posts that their close connections shared.
“So we developed and introduced a Feed that ranked posts based on what you care about most. Each part of the app – Feed, Explore, Reels – uses its own algorithm tailored to how people use it. People tend to look for their closest friends in Stories, but they want to discover something entirely new in Explore,” he said.
Ranking of content on Instagram
In a nutshell, Instagram has broken down how feeds and stories, Explore and Reels sections are ranked.
It said that for Feeds and Stories, the content from friends, families and close connections are prioritised in ranking and for this, several indicators based on the information about the post and the content itself, about the person who has posted it, the activity of the targeted user as well as the user’s history of interaction with the person are taken into account, after which the platform makes predictions regarding the ways the user might interact with a post. While Instagram has about 12 ways in which a person might interact with a post, the platform looks at five of them closely – the likeliness of a user spending a few seconds on the post, likeliness of the user commenting on it, liking it, saving it or tapping on the profile picture of the person who posted.
With this prediction, the platform ranks those posts higher in the feed or those stories first that users are more likely to interact with.
Mosseri also tried to showcase Instagram’s “willingness” to adapt by citing an example of how, based on earlier user feedback, stories that were re-shared from feed were ranked lower, but recently, with several issues, reshared posts were thought by people to be something that would amplify the reach of important content, and so Instagram modified its system.
Ranking of Explore and Reels is different from feed and stories, the company said, because it is where Instagram uses its systems to recommend content based on what it believes would interest the user. This is where users should be able to find new things, it said.
For this, Mosseri said that Instagram looks at the content which a user has liked, saved and commented on in the past, and then it looks for other users who have interacted with the same content. It then identifies the content that these other users interact with, presumably as a group, and then displays it to the user the next time they open Explore or Reels, “without us necessarily understanding what each post is about”. Usually, though, this system results in users seeing content related to the topics that they already like or follow.
The issue with Shadowbanning
In the blog post, Mosseri has also addressed the allegations of the platform shadowbanning or silencing people. He said that since the platform never explained why it took down any content that it does, people often reached their own conclusions about it, ultimately feeling confused or victimised. This, he said, was something the platform was working on. Mosseri said that the posts are taken down if they violate the Community Guidelines that the platform follows. Apart from this, if any user shares content that has been classified as ‘misinformation’ by third-party fact checkers, then it does not take this content down but does rank it lower. Moreover, content from users who repeatedly post content containing misinformation is made harder to find by the platform as a part of its initiative to reduce the spread of misinformation among the users.
He also added that some people equated getting fewer likes or comments as being shadowbanned, which was not true because most of the followers might not even see when a post has been shared since users do not scroll through more than half of their feed.
Managing the content you see
Instagram’s system takes into account the close friends of a user, placing their feed and stories at a higher rank, and also looks at the accounts that the user has muted. Apart from this, recommendations shown in feed, Explore or Reels can be modified by selecting the “not interested” option, so that the system does not show similar content in the future.
While Instagram has shared this post, making it out to be some major revelation, it does not seem to have brought to light any new information. For the most part, the factors that seemingly impact the feed, stories, explore or reels sections of the app were known to the users, and Mosseri has just given a bit more details about this.
However, the biggest concern that Instagram wanted to address with this post was the allegations of shadowbanning. While some insight has been given about silencing or hiding of posts containing misinformation, not much explanation has been given by the platform as to how it decides if some content violates its community guidelines. At present, the process of deciding whether any content violates the guidelines seems arbitrary at best, and without more information about such processes, it is not clear how much of this issue the app would be able to address.