Facebook’s news feed algorithm, which decides what content users see in their feeds, has gone through many changes over the years. The goal behind these changes has been to improve the user experience and show users content that is most relevant to them.
What is the Facebook news feed?
The Facebook news feed is the constantly updating list of stories, posts, videos, photos, and more that users see when they login to Facebook. It shows content posted by friends, pages, groups, and other connections. The news feed is unique to each user, tailored based on interests, connections, and interactions.
The news feed algorithm uses machine learning and AI to decide the ranking of posts. It takes into account thousands of factors to predict what content a user is most likely to engage with at the top of their feed. Some of the factors include:
- How recently content was posted
- How much engagement posts have received
- The amount of time a user spends interacting with certain connections
- How often a user likes, comments, or shares certain post types
- Which posts a user has interacted with in the past
By constantly monitoring user behavior and preferences, the algorithm can refine rankings to become more personalized over time.
Major changes to the news feed algorithm
Here are some of the major changes Facebook has made to its news feed algorithm over the years:
2009: Introduction of the news feed algorithm
When Facebook first launched in 2004, posts from connections were displayed in chronological order on a user’s homepage. In 2009, Facebook introduced a major update that implemented an algorithmic news feed ranking for the first time. This allowed the most relevant content to appear at the top rather than strictly recent posts.
2011: Creation of EdgeRank algorithm
Facebook developed its EdgeRank algorithm which incorporated more sophisticated machine learning to analyze thousands of factors about users and posts. It improved personalization and relevance of the news feed content.
2013: Incorporation of user surveys
Facebook began conducting surveys of users to get direct feedback about their preferences on what content they wanted to see more or less of. This explicit data improved the accuracy of the ranking algorithm.
2015: Friends and family focus
Facebook made changes to show more personal posts from friends and family and fewer public posts from pages, celebrities, and brands. The goal was to make the news feed feel more intimate and connected.
2016: Fight against clickbait
Facebook tweaked the algorithm to display fewer clickbait headlines, sensational content, and misleading posts in order to improve the quality of news feed articles.
2017: Meaningful social interactions
Facebook introduced updates aimed at promoting meaningful social interactions and connections between family and friends. Posts that sparked conversations were ranked higher than passive content.
2018: News feed ranking overhaul
Mark Zuckerberg announced a major change to prioritize “meaningful social interactions” over relevant posts. Content from friends and family would be ranked higher than public content even if less interesting.
2021: War against clickbait continues
Facebook made further updates to reduce clickbait and sensational content, emphasizing original reporting and content from friends.
2022: Promotion of original content creators
Facebook made changes to its algorithm to highlight more content from original creators and recommendations from friends rather than reshared content.
The impact of news feed algorithm changes
Facebook’s regular adjustments to its news feed algorithm have aimed to improve the user experience. However, the changes often receive backlash from certain groups:
- Businesses and publishers rely on Facebook for distribution and traffic. When public content is demoted, it hurts their reach and revenue.
- Political groups accuse Facebook of bias depending on which content gets promoted or demoted.
- Bad actors such as spammers and clickbait publishers try to “game” each algorithm update.
- Users accuse Facebook of censorship if they feel some content is unfairly demoted.
Despite the controversies, Facebook maintains that the changes create a better experience aligned with user preferences and meaningful engagement on the platform.
News feed algorithm transparency
To provide more transparency into news feed changes, Facebook launched its News Feed FYI blog in 2014. Major updates and modifications to the ranking algorithm are explained in detail.
Some examples of insights provided include:
- When Facebook prioritized original reporting in news posts, it explained the new signals used to identify original and republished content.
- For its 2021 clickbait update, Facebook outlined exactly how it detects clickbait patterns to train its machine learning models.
- When Facebook made changes to its content review process, it published details on its new methodology.
Despite providing the public with more information than ever before, Facebook is still reluctant to reveal the most sensitive inner workings of its algorithm that could be exploited.
Conclusion
Facebook’s news feed algorithm has gone through iteration after iteration over the years as the company tries to balance relevance, engagement, and meaningful connections. Major shifts have prioritized content in different ways, from recency to clicks to friend connections.
The impact of news feed changes is far-reaching for businesses, political groups, publishers, and users who rely on Facebook for distribution. While controversial, Facebook maintains the updates create a better user experience overall.
By publishing some insights through its News Feed FYI blog, Facebook does seem to be taking steps, albeit slowly, toward more algorithmic transparency. But the inherent complexity and business sensitivities mean we’re unlikely to ever know all the secrets that go into tailoring each person’s unique news feed.
Key news feed algorithm changes timeline
Here is a summary timeline of some of Facebook’s major news feed ranking updates:
| Year | News Feed Update |
|---|---|
| 2009 | Introduction of algorithmic news feed ranking |
| 2011 | Launch of EdgeRank algorithm |
| 2013 | Incorporation of user surveys for explicit feedback |
| 2015 | Prioritization of friends and family content |
| 2016 | Fighting clickbait and low quality posts |
| 2017 | Focus on meaningful social interactions |
| 2018 | Meaningful connections over passive consumption |
| 2021 | Continued demotion of clickbait |
| 2022 | Promotion of original content creators |
News feed ranking factors
Here are some of the key signals that Facebook’s news feed algorithm analyzes to determine rankings:
| Factor | Description |
|---|---|
| Recency | When a post was published |
| Popularity | How many likes, comments, and shares a post receives |
| Connections | How closely connected a user is with the person who posted |
| Interests | Whether a post is related to topics a user often engages with |
| History | What types of posts a user has interacted with in the past |
| Survey Data | Direct user feedback on what they want to see more or less of |
| Watch Time | How long a user spends viewing or interacting with video posts |
| Originality | Whether a post contains original content vs. repurposed or copied content |
Types of Facebook news feed content
The Facebook news feed algorithm analyzes and ranks many different types of content. Some common examples include:
| Content Type | Description |
|---|---|
| Status Updates | Personal posts shared by friends and connections |
| Photos | Images posted by users |
| Videos | Video clips, Facebook Watch shows, live streams |
| Articles | News articles, blog posts, links shared by users |
| Stories | Disappearing photo and video stories |
| Pages | Posts from followed pages of brands, celebrities, media |
| Groups | Posts in joined Facebook groups |
| Ads | Paid ads targeted to users based on data |
| Events | Upcoming event details |
| Marketplace | Listings for nearby items for sale |
News feed algorithm ML architecture
Facebook’s news feed machine learning system can be broken down into three core components:
Data ingestion
The system pulls in massive amounts of data in real-time about users, their connections, entities on Facebook, and existing posts.
Feature extraction
Important signals are extracted from the raw data that will be used as inputs for predicting engagement. These include features like post topics, user interests, post authors, etc.
Feed ranking
A deep learning model analyzes the extracted features to score and rank every post for a user. The highest scoring posts are served at the top of the news feed.
The ranking model is trained on historical engagement data and updated regularly to optimize for most relevant engagement patterns.
Fighting fake news in the news feed
Reducing the spread of false news and misinformation is a major challenge faced by Facebook’s news feed algorithm team. Some of their strategies include:
- Using ML to detect likely false news and send for fact checker review
- Lowering the ranking of confirmed fake news stories
- Cracking down on financially motivated spammers that spread misinformation
- Using signals like site reputation, click patterns, shared IP addresses, and account information to identify likely fake news domains and accounts
- Applying image analysis to detect manipulated media
- Leveraging user surveys and feedback to understand perceptions around misinformation
Despite these efforts, combating misinformation remains an uphill battle as malicious actors find new ways to exploit the news feed algorithm.
News feed bias controversies
Facebook’s news feed algorithm has faced criticism at times for perceived bias in how content is ranked and promoted:
2016 allegations of anti-conservative bias
Some conservatives accused Facebook of intentionally suppressing conservative news stories and topics after some anecdotal reports of reduced traffic. Facebook denied any intentional political bias.
Russia exploitation in 2016 US election
The news feed algorithm was manipulated by Russian trolls to spread misinformation during the 2016 US presidential election based on its engagement signals.
Canada engineering change
A 2019 report found that Facebook made an engineering change affecting Canadian users that incidentally increased exposure to conservative news vs. liberal.
Facebook maintains neutrality
Facebook maintains that despite isolated incidents, the news feed algorithm framework itself remains non-biased. The company is committed to fairness and neutrality in how content is ranked.
The future of Facebook’s news feed
Looking ahead, we can expect Facebook’s news feed algorithm to keep evolving as new technologies emerge and new challenges arise, including:
- Integrating new forms of user feedback such as comments or reactions to improve personalization
- Incorporating multimedia analysis of images, videos, and live streams to understand content
- Identifying harmful viral misinformation early before it spreads widely
- Promoting more local content as Facebook expands into international markets
- Balancing user privacy controls with personalization
Changes to the news feed algorithm will remain controversial, with groups attempted to steer rankings in their favor. But the core goal will continue to be keeping users engaged by surface relevant, high-quality content from their network.