Recently, Meta has launched 22 different explanatory cards to provide users with specific explanations on how the company uses AI to control the content they see on Instagram and Facebook.
Meta states that this move aims to provide more transparency to its AI system, but the specific explanations on these cards still remind users that they live within algorithms and are precisely manipulated by algorithms.
Here are some of the 22 explanatory cards from Meta:
<img src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/5a27b544-39e9-43d2-9aad-6201f05e0c7f.png?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg" alt="Facebook explanatory card" original-src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/5a27b544-39e9-43d2-9aad-6201f05e0c7f.png"/> Explanatory card for Facebook
<img src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/ea92ccd0-4e45-43a6-ab47-55040d9a061c.png?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg" alt="Instagram explanatory card" original-src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/ea92ccd0-4e45-43a6-ab47-55040d9a061c.png"/> Explanatory card for Instagram, Source: Meta Transparency Center
Each card provides detailed and easily understandable information, explaining how the AI system behind these features ranks and recommends content.
Here are a few examples:
<h2>How Facebook Reels show content to users</h2>
When users browse and interact on Facebook, an underlying AI system provides Reels (short videos), including content from creators that users may be interested in but have not yet followed, as well as cross-app recommended content from Instagram.
The AI system that powers Facebook Reels automatically determines which Reels to show to users and in what order, based on predictions of what content users are most likely to be interested in or engage with. These predictions are based on various factors, including users' recent follows, likes, and interactions with other users and content. Here's how it works:
<blockquote>
1. Collect inventory information
First, the AI system collects all the Reels that users may be interested in, which may include Reels from users or accounts they follow, or Reels similar to those they have recently interacted with. The AI system may also recommend Reels from sources similar to users' followed or interacted accounts.
2. Utilize signaling cues
Next, the AI system takes into account various cues related to each Reel, such as its duration, similarity to other Reels, and relevance to the content users are willing to engage with, and runs a simple model to select approximately 10-100 most relevant Reels (problematic content is filtered out).
3. Make predictions
At this stage, the AI system utilizes models to help predict the most relevant and valuable content that users would find.
4. Sorting Reels by Score
Lastly, the system calculates the relevance for approximately 200 posts and sorts them by score. The system predicts that Reels with higher value will be displayed at the top of the feed.
Meanwhile, users can reduce or increase the display of similar Reels by hiding or saving and sharing them.
<h2>How Facebook Stories Show Content to Users</h2>
According to the information card, Facebook Stories (a feature that allows users to post photos and videos that disappear after 24 hours) operate as follows:
1. Collecting Stories
First, the system collects all relevant Stories shared by users or public pages within the past 24 hours (excluding any non-compliant Stories).
2. Prediction and Analysis
Next, the system gathers all Stories that can be shown to specific users and predicts which ones they would find most relevant and valuable. The system keeps these Stories and removes the rest.
Factors considered by the system include: the number of times users click to view the Story in full screen, the total duration of browsing the author's Stories, the number of different authors' Stories and Story sets viewed, the number of times users respond to the author's Stories through likes or chats, the total duration of browsing the author's Stories, and the average time spent browsing Stories.
3. Sorting Stories
The system then sorts this small portion of Stories based on the likelihood of user interaction with each Story.
4. Applying Additional Rules
Finally, the system applies additional rules to ensure a balanced display of Stories from users and public pages.
Users can share Stories with others through Messenger, add them to their own Stories or public page Stories, or choose not to view Stories. If users choose not to view Stories, they will not see any other Stories posted by the creator of that Story or any other public page, unless they choose to view them again.
<h2>How Instagram Explore Shows Content to Users</h2>
The system ranks content based on factors such as the time since the post was published, the total number of times users view or click on the post thumbnail, the number of times the post is prioritized in a series of posts, and the number of times users click &#34;Not Interested&#34; on the post. Additionally, the system considers the number of times users browse posts from authors with similar interests to the author of the post.
Finally, the content is ranked. The system predicts that it will provide users with higher-value content and push it to the top of the &#34;Explore&#34; tab.
Similarly, users can influence this process by saving content or marking it as &#34;Not Interested&#34; to encourage the system to continue showing or filtering similar content in the future.
Users can also view Reels and photos that the algorithm did not specifically select for them by selecting &#34;Non-Personalized&#34; in the Explore filter.
</blockquote>
<h2>How Instagram Search Displays Content to Users</h2>
When users view content and engage on Instagram, an underlying AI system provides results when users search for content.
The AI system that Instagram Search relies on automatically ranks search results by predicting the most relevant and valuable content to users.
Here's how it works:
<blockquote>
1. Collect inventory information.
First, the system collects all relevant search results to rank them for users. This can include topic tags, locations, Reels, posts, profiles, audio, or other results related to the words users search for.
2. Score the results.
The system then scores each search result based on various factors, such as the type of content and its relevance to the content users typically engage with. Factors considered include:
Similarity between the words used in the search and the words used in the account username or profile name, similarity between the words used in the search and the words used in suggested keywords, similarity between the words used in the search and the words used in topic tags, and the number of times users in the same country/region who performed the same search clicked on topic tags.
3. Apply additional filters.
The system applies &#34;additional filters&#34; and the &#34;integrity process&#34; to narrow down the range of relevant search results for users.
4. Sort the results by score.
Finally, the system prioritizes and displays the results predicted to be most valuable or relevant to users.
</blockquote>
At the same time, the system customizes the Instagram search experience based on users' dynamics. Users can choose to control or customize the displayed content and also view non-personalized search results.
<h2>Can users counteract?</h2>
Regardless, in the era where everyone lives in algorithms, Meta's approach to increasing algorithm transparency is commendable.
Research shows that algorithm transparency can play a role in accountability and the right to know.
First, algorithm transparency makes algorithm operators more accountable. If there are biases in accuracy and fairness, the disclosed algorithm can be used to hold algorithm operators responsible.
Second, algorithm transparency gives algorithm-regulated entities a certain degree of right to know. This right to know is beneficial for third parties, especially professionals, to supervise and for algorithm-regulated entities to question the fairness and rationality of algorithmic decisions based on the disclosed algorithm after the fact.
Of course, Meta's release of this information is also in line with regulatory trends. Currently, European legislators are rapidly advancing legislation to propose new interpretations and transparency requirements for companies using AI technology. US legislators have also expressed their desire to begin drafting similar legislation later this year.
In addition to the 22 cards already published, Meta also stated that it will expand the scope of explanations to include features such as &#34;Why am I seeing this post&#34; in the coming weeks.
Furthermore, Meta offers a feature that allows users to have centralized control over the content they want to see on Facebook and Instagram.
Instagram already supports selecting &#34;Not Interested&#34; for certain posts to increase recommendations of content that is less similar. Soon, users will also be able to select &#34;Interested&#34; to view certain types of content, and in the future, Meta will provide even more diverse options.
Over time, users will be able to counter AI to some extent as well.
<img src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/d3283b1c-8001-44ce-b996-76e0b709ca2a.png?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg" alt="Image" original-src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/d3283b1c-8001-44ce-b996-76e0b709ca2a.png"/>

Living in Algorithms? Meta's Heavyweight Release: Algorithms Precisely Grasp You