Meta has recently released a blog post detailing the social media algorithms that power its platforms, Instagram and Facebook. The company aims to provide transparency and accountability by demystifying how content is recommended to users. In the blog post, Meta’s President of Global Affairs, Nick Clegg, emphasizes the company’s commitment to openness in the face of concerns surrounding advanced technologies like generative AI.
The blog post introduces 22 “system cards” that explain how content is ranked and recommended for users on Facebook and Instagram. These cards cover various features, including the Feed, Stories, and Reels, and provide detailed yet understandable information about the AI systems behind them. For instance, the system card for Instagram Explore, a feature that displays content from accounts users don’t follow, explains the three-step process utilized by the AI system.
To allow users to have more control over the content they see, Meta includes information on how users can influence the recommendation process. Users can save content to indicate preferences and prompt the system to show similar content. Additionally, marking content as “not interested” helps the system filter out similar content in the future. By selecting the “Not personalized” option in the Explore filter, users can view reels and photos that haven’t been specifically tailored by the algorithm. More comprehensive information about Meta’s predictive AI models, input signals, and frequency of content ranking can be found in the Transparency Center.
The blog post also highlights other features that inform users about the content they see and allow for further customization. The “Why Am I Seeing This?” feature is being expanded to Facebook Reels, Instagram Reels, and Instagram’s Explore tab, enabling users to understand how their previous activity influenced the system’s content suggestions. Instagram is also testing a “Interested” feature for recommended reels, allowing users to mark content they find appealing for future similar recommendations. The ability to mark content as “Not Interested” has been available since 2021.
Related Post
In addition to providing more insights into its algorithms, Meta is rolling out its Content Library and API, a suite of research tools containing public data from Instagram and Facebook. These tools will offer researchers the opportunity to search, explore, and filter data, with access granted through approved partners. The University of Michigan’s Inter-university Consortium for Political and Social Research will be among the first partners. Meta believes these tools will provide the most comprehensive access to publicly available content across the platforms, fulfilling its data-sharing and transparency obligations.
The motivation behind Meta’s efforts to explain its AI algorithms stems partly from the increasing regulatory scrutiny surrounding AI technology. As AI advances and gains popularity, regulators worldwide have expressed concerns about data collection, management, and usage. Although Meta’s algorithms are not new, the company’s mishandling of user data during the Cambridge Analytica scandal serves as a reminder to prioritize transparent communication.
By publishing detailed information about its algorithms and providing users with more control over their content preferences, Meta aims to establish trust and address concerns regarding the impact of AI on individuals’ online experiences. Through its commitment to transparency and openness, Meta hopes to alleviate worries and foster a better understanding of how its platforms operate.
