Meta is publishing new content to explain how its recommendation engine works across different products. As part of that, it is also rolling out a new option called "Why am I seeing this?" for Instagram and Facebook Reels as well as Instagram's Explore page. In this way, you'll be able to understand why that post is being shown to you.

In a blog, the president of Global Affairs at the company, Nick Clegg, explained how the company wanted to become more transparent about its AI systems and give users greater control over the content that is shown to them.

"Today, we're building on that commitment by being more transparent about several of the AI systems that incorporate your feedback in ranking content across Facebook and Instagram. These systems are more likely to make you see posts that are actually relevant and interesting to you, Clegg said.

Meta is trying a new system on Instagram: posts will now tag the "Interested" tag on you so that you may see more of that content. As it is right now, users can tap the three-dot menu within the post and then select "Not Interested" in order to see fewer of that sort of content.

The company said it is also working on making Facebook's "Show more, Show less" controls "more prominent." Today, you can find them in the three-dot menu on a recommended Reel or a post. It's unclear if Meta is relocating these controls elsewhere.

Today, the company published 22 "System Cards" explaining how many different systems work within its apps, from the reasoning behind Facebook Feed recommendations to Facebook Group timelines, suggested people and groups, Instagram Reels recommendations, the notifications system, and Stories AI. These cards extend on these topics by means of examples but could be a long read to go through and understand for a normal user.

According to Clegg, the company has come out in the Transparency Center with information on some of the signals that influence his/her recommendation — such as the number of likes or shares received on a post. Yet, the company has stayed out of "disclosing signals which could potentially make it easier for people to circumvent our defenses."

For one, Meta has weighed in on the possibility of building models at a scale of "tens of trillions" of parameters, hundreds of times bigger than even current models such as ChatGPT and GPT-4. My colleague Devin Coldewey in his story astutely observes that such huge models have a lot to say, including churning through volumes of data insights on a particular user.

"Think about it: an AI model that is as big or larger than any yet built. What goes in one end is every action you undertake on Meta's platforms and what comes out the other is a prediction of exactly what you will do, or like, next. Kind of creepy, is it not?"

This therefore means that the user controls of some other contents couldn't do much to regulate some of the content that is being given as recommendation. In 2021, Clegg posted on Medium where he posited that people's actions are as good as the algorithms in rendering different contents on the feed. But users do not know what all the signals are captured by a particular platform so that it could train its AI-powered feeds.

Meta is also opening its algorithms to research through a new Content Library and API, which features public posts, pages, groups, and events on Facebook, as well as public posts and data from creators and business accounts on Instagram.

Meta has taken heavy criticism the past few years for how it serves content recommendations to users. Last year, Facebook whistleblower Frances Haugen complained the algorithms of the company had "unhealthy side effects on important slices of public content, like politics and news," plus Instagram's damaging effects on teens. In a separate development last year, Meta had to settle with the U.S. Justice Department over its ad-serving algorithms allegedly depriving users of housing opportunities.

Instagram has also been increasing AI-driven content in the feed. According to the Q1 2023 earnings call, Mark Zuckerberg reportedly stated that time spent on Instagram jumped by 24% because of AI-aided Reels suggestions. Meta CEO said that the company wants to increase such content by filling up 30% of the feed with algorithmic recommendations. The moves come amid a string of attempts by Meta to address the concerns of regulators and privacy advocates ahead of rolling out even more AI-picked content.

Since there has been the rise of decentralized networks such as Bluesky and Mastodon, it would be interesting to see lots of conversations in the regard to having control over some of the algorithm that would drive your feed. Thus, platforms operated centrally such as Instagram and Facebook could learn a lot from those other social media sites about better algorithmic choices for users and some form of control.