Over the past year, we’ve strengthened the default protections behind whatInstagram recommends in India to reduce unwanted content and improve people’sexperience on Instagram. This work has been guided by what our users told us theywant from their recommendations.
Today, we’re sharing an update on this work.
How Recommendations Work
At Instagram, we make personalized recommendations to help people discover newcommunities and content. Our goal is to make recommendations that are relevantand valuable to each person who sees them. For example, if you interact withrestaurants and bookstores on Instagram, we may recommend content about food,recipes, books, or reading.
Recommendations include content from accounts you don’t already follow, in placeslike Explore, Reels and Feed, so we apply guidelines specific to them. Through ourRecommendations Guidelines, we work to avoid recommending content that couldbe low quality, objectionable or particularly sensitive. We use technology to detectcontent and accounts that don’t meet those guidelines and to help us avoidrecommending them.
As always, content that goes against our Community Standards is removed fromInstagram, and content that is unlawful in a country is restricted there.
What We’ve Changed
Better detection: Around the world, our systems check content against ourRecommendations Guidelines before we recommend it. We’ve been improving thosesystems, including using more capable AI models that assess content directly ratherthan waiting for someone to report or hide it, so we catch more of what people don’twant to see.
Tuned for India: We’ve improved our systems specifically for India so they catch moreof what people here told us they didn’t want to see. Today, our systems keep more ofthis content out of recommendations, including sexually suggestive content andcontent that depicts violence.
Since these changes, we’ve been able to proactively reduce more content in Indiathat doesn’t meet our Recommendations Guidelines. As a result, we’ve seen peoplereport recommended content about half as often as they were a year ago.
Your feedback shapes this too. When someone hides a post or taps “Not interested,”that tells us what not to recommend to them, and helps us understand what peopledon’t want to see.
Controls Over What You See and Who Can Reach You
Alongside the standards we apply for everyone, you have controls to adjust what yousee and who can reach you:
·
Sensitive Content Control lets people choose to see less sensitive content in theirrecommendations, across Search, Reels, Explore, suggested accounts, and in-Feedrecommendations. It’s in Settings and Activity, under Content Preferences. In India, themost mature content isn’t recommended, so this control adjusts what you see withinthat.
·
Political Content Control lets people choose to see less political content inrecommendations from accounts they don’t follow. It’s in Settings and Activity, underContent Preferences → Political Content.
·
Not Interested lets people tell us a recommendation isn’t for them, by tapping the threedots on a post or reel.
·
Hidden Words filters chosen words, phrases, and emojis out of comments and messagerequests.
·
Block prevents someone from finding a profile, seeing posts, or sending messages, andalso covers new accounts they try to create.
Teens have additional age appropriate protections. Last year, we introduced TeenAccounts in India, which automatically places teens in a 13+ content setting shapedby our age-appropriate content guidelines.
Our Continued Work
We’re committed to improving our systems as user behavior and technologychanges, and we’ll keep listening to what people in India tell us they want to see.
We publish data on how we enforce our policies in our global Community StandardsEnforcement Report, and in the monthly transparency report we publish for Indiaunder the IT Rules, 2021.