If you’ve ever used Instagram, you’re probably very familiar with one tab — maybe the best part of the the Facebook-owned app. This feature takes users to content Instagram thinks might catch your interest, but isn’t included in the endless scroll of photos and Stories from the same several hundred people and brands you follow.
In a blog post published Monday, Instagram toued new information on how this feature works. We’re talking about, of course, the Explore feature.
Instagram claims over half of its users utilize the Explore tab. To put that in perspective, Instagram has over 1 billion users. You might think the Explore tab runs using a simple algorithm that looks at what pages you follow and finds similar pages other people like, but Instagram has revealed that the brains behind the Explore tab are a lot smarter than that.
“Content-based models have difficulty grasping such a variety of interest-based communities”
“People publicly share billions of high quality pieces of media on Instagram that are eligible inventory for Explore. It’s challenging to maintain a clear and ever-evolving catalog-style taxonomy for the large variety of interest communities on Explore — with topics varying from Arabic calligraphy to model trains to slime,” the blog post reads. “As a result, content-based models have difficulty grasping such a variety of interest-based communities.”
Accounts > Individual Posts
To deal with this problem, Explore focuses on showing you accounts you may enjoy, rather than individual posts you might like. Explore utilizes a type of machine learning called “word embedding” to find pages you might want to check out. Essentially, it locates words in post captions that are often near each other to determine what kind of content these accounts are posting.
If you put the words “music” and “festival” together a lot, then you’re probably posting similar content to other accounts that also have those words together. That’s part of how Instagram selects content for you.
“Evaluating even just 500 media pieces through a deep neural network for every scrolling action requires a large amount of resources,” the post reads. “And yet the more posts we evaluate for each user, the higher the possibility we have of finding the best, most personalized media from their inventory.”
Instagram looks at accounts you’ve interacted with in the past, which might just mean you liked some posts, to start the process. The company refers to these as “seed accounts.” Instagram’s A.I. then locates accounts that appear to be related to pages you already like using its embedding techniques. From there, the system uses a “complex ranking model” to identify the best content to show you. The more you interact with different pages, the better Instagram gets at finding you more pages you’ll like.
How you interact with media is taken into account in different ways. Instagram might think you saving a post is more significant than you liking a post, so it will put more weight in you making that decision. If you say you want to see less content like a post you interacted with, that’s factored in as well.
“We also want Explore to be a place where people can discover a rich balance of both new interests alongside existing interests,” the post reads. “We add a simple heuristic rule into value model to boost the diversity of content.”
Instagram claims it’s still working to make the Explore feature even more advanced and useful. The company says Explore is “continuously evolving.”
As long as I keep getting my corgi videos, I’ll be happy.