On Thursday (29), Meta published on its official website an explanation of the operation of the algorithms of its social networks – Facebook and Instagram. In the post, Mark Zuckerberg’s company gives a basis of how recommendations appear to users, which is a big mystery to some people.
In 2021, Meta also disclosed how suggestions were made in Instagram’s Feed and Explore. However, the artificial intelligence (or algorithm) used by social networks is constantly evolving. These updates are performed to improve the user experience and revamp content based on people’s tastes. After all, human beings change: what I can enjoy today may not be the same a year from now—or less.
Interactions on Facebook and Instagram are influenced by friends
You know that mother’s talk of “that friend of yours is a good/bad influence”? This is one of the best definitions about the recommendation method of Facebook and Instagram. In the Facebook section, Meta explains that the priority is your “virtual circle”: posts shared by friends, the content of pages you follow, and groups you’re a part of.
Then, artificial intelligence evaluates the “input signals,” such as the history of interaction with the authors of a publication. The algorithm selects the contents based on the relevance score acquired from these signals. The posts with the highest scores lead the “ranking” and are prioritized in your Feed.
On Instagram, the operation does not differ much. There is an evaluation of the interaction with the accounts and signals to recommend content in the feed, Explore and Reels. For this social network, Meta also presented how the algorithm prioritizes the order of stories.
In addition to the history of interaction with the author, the artificial intelligence evaluates how you react with the story (whether you respond, like or return to view) and the likelihood of meeting the author in person – now Bruna Biancardi can get a sense of why some models appear on the list of Neymar’s first stories.
Meta teaches how to “train” the algorithm and use resources to do so
Among the 22 cards about the workings of the algorithm of your social networks and on your blog, Meta also teaches you how to improve recommendations and new features to deliver content more consistent with the user’s taste.
It is not difficult to train the algorithm. One of the methods is to interact with themes you like (for example, my Feed and Explore is filled with videos from The Office series). Meta also recommends hiding or unfollowing what’s irrelevant to you, managing bookmarks, reporting inappropriate content, and clicking the “Show more” or “Show less” buttons—both of which will be “renewed” to make them more accessible to users.
Meta teaching users how to train the algorithm to improve their experience with the app resembles a TikTok update. In March, ByteDance’s social network launched a tool to “reset” the recommendation algorithm.
Reason for disclosing was not “goodwill”
Meta released the information from its algorithm not out of kindness, but to advance on European Union legislation. From 2024, the Digital Markets Act comes into force in EU member countries. So tech and online services companies will have to act more transparently about their algorithms.
Meta Verified does not impact the algorithm
Last week, Meta launched Meta Verified, a subscription service for account verification. At launch—and in the publication reporting how the algorithm works—the company did not say whether Meta Verified impacts the recommendation of content. That is, an account with the purchased blue badge will not have priority in the display ranking in the feed.
