A new blog post from Meta has finally explained how the elusive algorithm works behind its apps Facebook and Instagram.
Nick Clegg, the President of Global Affairs, has published a blog on Meta’s website elaborating on how the algorithm behind its social media platforms Instagram and Facebook works.
The algorithm is designed to surface similar content that you’ve already been viewing. However, for years it has remained shrouded in mystery just how it operates. However, in a bid to be more transparent with its users, Clegg and his team have decided to explain how the AI systems work.
Essentially, the core point is that the AI will “predict how valuable” content is to the user, and then pushes it up. It then predicts what you’d like based on sharing, length of time watched, and more.
Clegg goes on to explain that a wide array of AI predictions are in place to serve content to the user, but that it isn’t a “perfect gauge”. To rectify this, Meta is publishing what it calls “system cards”.
The system cards give details on just how the algorithm sees you, including content that you might have been recommended by creators you’ve never expressly sought out. Included in this release is what Meta calls “signals” which is what the AI uses to determine which content to bring to your screen.
The ex-Deputy Prime Minister also details that Meta is aware that users might not go digging through the website to find these system cards. To counter this, they’ve made them available in the apps themselves.
Currently, if you press the three-dot icon on a “suggested post” on Instagram, and choose “Why am I seeing this?” you get a vague response. Clegg has stated that Meta intends to expand this to include far more information.
The company is also working to include a feature similar to TikTok’s “Not interested” button for Instagram. This tells the algorithm that you don’t want to see content similar to what’s being shown and will alter the feed for you.
Clegg also mentions that they intend to make these buttons more visible to users on Facebook as well.
Meta is also committing to helping researchers with new tools. As the company slowly pivots to being a lot more transparent in how it operates, a public Content Library and API will be made available. This contains all the data necessary to fully understand Meta’s internal systems.