liveBook · Manning

Mutual influence

As explained in Hidden Influences

A reciprocal causal relationship in which users affect algorithms and algorithms affect users.

Mutual influence is a feedback loop: users affect what algorithms learn, while algorithmic recommendations affect what users encounter and do. Following an unfamiliar person can produce more recommendations from that person, making further interaction more likely. section 1.3.3

How recommendations become more extreme

Suppose users interact with only 10% of posts. As the network grows, competition for attention increases, encouraging emotionally charged language that performs well. Recommendations then shift toward more extreme content. Similar feedback can help fringe online groups attract followers and gain offline influence, illustrated by the trajectory from Gamergate to the January 6, 2021, attack on the US Congress. section 1.3.3

How Hidden Influences covers it

How deep it goes
Traces a temporal feedback loop from interaction data to recommendations and increasingly extreme content, without implementation details.
Examples
Uses follows, recommendation changes, a 10% interaction scenario, Gamergate, and January 6, 2021, as concrete illustrations.
What you can do after
Evaluate how interaction data can reinforce recommendations and increase the prominence of emotionally charged content.
How it connects
Connects recommender systems with user preferences, social media, online communities, and emotionally charged content.

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.