Meta explains how AI predictions influence Facebook and Instagram recommendations

June 29 (Portaltic/EP) –

Goal has explained how Artificial Intelligence (AI) predictions influence Facebook and Instagram recommendations, with the publication of 22 new System Cards, cards that offer information on how its algorithm classifies content.

The company announced a few days ago a community forum dedicated to generative AI, a space for deliberation with which they seek to determine the principles that will guide the design of your products based on said technology.

The President of Global Affairs, Nick Clegg, He indicated then that as Meta refined the best ways to develop generative AI systems, this community forum would allow the company to learn “how people would like models to behave for nuanced issues and therefore inform future policy and product considerations”.

With the aim of reinforcing his commitment to transparency in his services, the manager has now delved into the way in which that influence AI predictions in the recommendations it offers both on Instagram and Facebook.

First, it has recognized that “sharing a post is often an indicator” that a post has been found interesting by the user, so predicting that they will share it is a factor that their systems will take into account, but not the only one.

“We use a wide variety of predictions combined to get as close as possible to the right content, including some based on behavior and some based on user feedback. received through surveys”, has pointed out.

With this, he has remembered that they use the calls System Cardsthat allow to know how their systems work in an accessible way “for those who do not have deep technical knowledge” and that now it has 22 new ones.

These cards offer information about how AI systems classify content that may be most relevant to users, as well as controls that can be used to personalize the experience.

The System Cards they also include explanations about the recommendations in the ‘feed’, the Stories and the ‘reels’, and integrate AI systems that recommend ‘not connected’ content, that is, from people or profiles that are not followed.

As a complement to the System Cards, Meta has said that it also has types of tickets (also called ‘signals’), which are used in the classification of the ‘feed’ of Facebook and that are also found in the company’s Transparency Center.

The technology has also commented that it uses these signals to identify harmful content, which it deletes as soon as it is “aware of its existence.” He has also pointed out that uses them to reduce the distribution of other types of content problematic according to the community guidelines.

On the other hand, Clegg has commented that he will soon expand the function called ‘Why am I seeing this?’ -available in the ads offered by their platforms- in the Instagram Reels tab and in Explore, as well as on Facebook Reels.

To maintain “an open approach to research and innovation,” the company will roll out a new set of tools for researchers in the coming weeks, such as the content library and application programming interface (API). English) of Meta.

The first includes data from public posts, pages, groups, and events on Facebook. Your data is searchable, explore and filter in a graphical user interface or through a programmatic API.