liveBook · Manning

Algorithmic feed

As explained in Hidden Influences

A feed whose contents or ordering are selected by algorithms rather than solely by chronological or follow-based rules.

A social media feed can contain more updates than anyone has time to read. An algorithmic feed helps select or order those updates, often using personalized relevance instead of showing friends’ posts solely in reverse-chronological order. It can also add content from people the user does not follow. section 1.1.1 section 1.2.2

From recent updates to relevant updates

Facebook’s early algorithmic feed used EdgeRank to weight posts and friends. Newer posts received priority, as did updates from friends a user interacted with more often. This simple weighting formula differs from the machine-learning algorithms used by platforms today. section 1.2.2

Interactions can also create a feedback loop: seeing more updates from a friend may lead someone to interact with that friend more, which in turn can make their updates more prominent. The system infers preferences from behavior rather than relying only on an explicit list of interests. section 1.2.2

Figure 1.3 Various social media models. In the Subscription model, the content is seen only by users who explicitly follow the content producer, with no options or resharing. In the Network model, users can also reshare content from users they follow, enabling the distribution of content outside of the immediate network. Finally, in the Algorithmic model, an algorithm can add content to users’ feeds, even with no direct connection to the content itself or its author.

Compare feeds limited to followed accounts with network resharing and algorithmic additions from outside a user’s connections. section 1.2.2

Is one source being amplified?

Suppose a reader follows two news outlets, but one supplies 80% of what they see. That imbalance might signal amplification, but it is not enough to establish it: the outlets may publish different amounts, and the reader’s interests and behavior may also matter. If one account posts once and another posts ten times, seeing all 11 posts produces a ten-to-one difference without algorithmic skew. section 5.5.3

Personalization can make content easier to find and bring people with shared interests together, while platforms can optimize for engagement or time spent. It can also contribute to filter bubbles and more uniform culture. Personalized feeds are difficult to audit from outside, raising the question of whether content is valuable to users or merely engaging. section 1.1.1 section 7.5.2

How Hidden Influences covers it

How deep it goes
Explains ranking weights and feedback loops, and develops feed-amplification comparisons without detailing modern machine-learning algorithms.
Examples
Uses Facebook’s EdgeRank and comparisons of two followed accounts with different posting rates to illustrate ranking and amplification.
What you can do after
Evaluate feed amplification by comparing content seen with content available, while considering user behavior and posting rates.
How it connects
Connects recommender systems with feedback loops, filter bubbles, and amplification.

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Hidden Influences. The section links point to the parts of the book it draws on.