Millions of individuals actively engage with Facebook and Instagram, using these platforms to share their experiences, connect with like-minded individuals, and discover content that resonates with them. To ensure that everyone’s experience on our applications is unique and tailored to their preferences, we utilize artificial intelligence systems that determine the content that appears, taking into account user choices.
Meta aims to challenge the misconception that algorithms render individuals powerless over the content they encounter on Facebook and Instagram. To address this, we acknowledge the need for greater transparency and control over the content users see.
Today, we are reinforcing our commitment to transparency by providing more clarity on several artificial intelligence systems that leverage user feedback to rank content across Facebook and Instagram. These systems increase the likelihood that the posts users come across are relevant and interesting to them. We are also making it easier for users to better control their content experience on our apps by testing new controls and enhancing accessibility. Additionally, we are sharing more detailed information with experts to facilitate their understanding and analysis of our systems.
This commitment to openness, transparency, and accountability is part of a broader ethos. As transformative technologies like generative artificial intelligence advance rapidly, it is natural for people to be both excited by the possibilities and concerned about the associated risks. We firmly believe that the best way to address these concerns is through transparency. Consequently, we believe that companies should be more forthcoming about how their systems operate and actively collaborate with industry, government, and civil society to ensure responsible development. It starts with providing users with greater insight and control over the content they encounter.
The Influence of Simulated Intelligence Predictions on Proposals
The impact of simulated intelligence forecasts on proposals is significant. Our AI systems predict the relevance of content to you, allowing us to show it to you in a more timely manner. For example, if you share a post, it often indicates that you found it interesting, so our systems take that into account when predicting if you will share a post. However, no single prediction is a perfect indicator of the relevance of a post to you. Therefore, we use a diverse range of predictions, including behavioral and user feedback-based predictions, to get as close as possible to accurate results.
We believe in being transparent about how our systems function. One way we have embraced transparency at Meta is through the publication of system cards, which provide insights into how our systems work in a way that is accessible to both technical and non-technical individuals. Today, we are releasing 22 system cards for Facebook and Instagram.
These system cards offer information about how our AI systems rank content, the predictions each system makes to determine relevant content, and the controls available to customize your experience. They cover various features such as Feeds, Stories, Reels, and other surfaces where people discover content from the accounts or individuals they follow. In addition to the system cards, you can find a more detailed explanation of the AI behind content recommendations on our website.
To provide further details beyond what is included in the system cards, we are sharing the types of data sources, known as signals, as well as the predictive models that these signals inform. These models play a crucial role in determining the content you see as most relevant from your network on Facebook. The categories of signals we are disclosing encompass the majority of signals currently used in Facebook’s content ranking. You can find these signals and predictions in the Transparency Center, along with information about how frequently they are used in the overall ranking process.
We also utilize signals to identify and remove harmful content, as well as to reduce the distribution of other types of harmful or low-quality content in accordance with our Content Distribution Guidelines.
While we are providing examples of the signals we use, there are limitations to what we can disclose for safety reasons. We want to be transparent about our efforts to keep bad content away from people’s Feeds, but we must also exercise caution to avoid revealing signals that could potentially enable individuals to bypass our protective measures.
We understand that not everyone will access this information on our website, which is why we enable users to view details directly within our applications. You can see why our systems predicted content as relevant to you, as well as the types of activities and data sources that may have influenced those predictions.
We are placing a particular emphasis on the Instagram Reels tab and Explore, and Facebook Reels in the near future, following our previous introduction of these features for some Channel content and all ads on both Facebook and Instagram. You will have the ability to tap on a specific reel to obtain more information about how your past actions may have informed the AI models that shape and deliver the reels you see.
Enhancing Personalized Experiences
By leveraging the available tools at your disposal, you have the ability to tailor your experiences on our applications, ensuring that you see more of the content you desire and less of what you don’t. To simplify this process, we have created centralized locations on Facebook and Instagram where you can modify controls that influence the content displayed on each platform. You can access your Channel Preferences on Facebook and the Recommended Content Control Center on Instagram through the three-dot menu on relevant posts or via the Settings menu.
On Instagram, we are currently testing a new feature that allows you to indicate your interest in a suggested reel within the Reels tab, enabling us to show you more of the content you enjoy. Since around 2021, the “Not Interested” feature has also been available. To learn more about influencing the content you see on Instagram, you can find additional information here.
To further customize your experience and the content you encounter, we offer a “Show more, show less” feature on Facebook, accessible through the three-dot menu on all posts in your Feed, Video, and Reels. We are actively working on making this feature even more prominent. Additionally, if you prefer not to have an algorithmically curated Feed or simply want to see how your Feed would look without it, you can utilize the Channels tab on Facebook or switch to the Following option on Instagram for a chronological Feed experience. Both on Facebook and Instagram, you also have the option to add accounts to your Favorites list, ensuring that you consistently see content from your preferred creators.
Enabling Scientists with Enhanced Resources
We strongly believe in fostering an open approach to research and development, particularly when it comes to groundbreaking artificial intelligence technologies. Instead of solely relying on a few major tech companies, we have been dedicated to releasing over 1,000 artificial intelligence models, libraries, and datasets for researchers over the past decade. Our aim is to provide scientists with access to our computing power, allowing them to pursue research directly and securely. As we move forward, we are committed to maintaining transparency and making more artificial intelligence models openly available in the future.
In the coming weeks, we will be introducing a new suite of tools specifically designed for researchers: the Meta Content Library and API. The Library encompasses data from public posts, pages, groups, and events on Facebook, while Instagram’s inclusion will consist of public posts and data from creator and business accounts. Researchers will have the ability to search, analyze, and filter the data from the Library through a user-friendly graphical interface or an automated API.
To ensure secure data sharing for research purposes, eligible scholars and research institutions pursuing scientific or public interest research topics will be able to apply for access to these tools. This process will be facilitated through partnerships with trusted organizations with expertise in secure data sharing, starting with the University of Michigan’s Inter-university Consortium for Political and Social Research. These tools will provide the most comprehensive access to publicly available content across Facebook and Instagram, surpassing any research tool we have previously developed. Moreover, they will help us meet new obligations concerning data sharing and transparency compliance.
By introducing these products to researchers early in the development cycle, our goal is to gather valuable feedback and ensure that we are building the best possible tools to meet their needs.
