Filter Bubbles

Last week, we looked at collaborative filtering and came to understand the power of estimating how we might fill in unknown entries in a matrix containing users’ ranking of content. In this question, we will consider some of the caveats associated with using these techniques. 4A) Consider a political issue with sides A and B. How would collaborative filtering respond to a user who tends to prefer content siding with A? Discuss how this may affect political campaigns. 4B) If a user gives a positive signal (say, a like on Facebook or a rating of 5 stars on some other platform (or even just continues to watch!)) to content from A, between a piece of content from A and a piece of content from B, which is the user more likely to see next? Then, assuming they give a positive signal to that, what happens to content in the future? 4C) Consider content from a large corporation that has resources to invest in advertising. As a result of a successful marketing campaign, their content has received positive reviews from 60% of customers. Meanwhile, a small content creator receives positive reviews from a smaller number of customers (say, 20%). Which would appear to be more appealing to a new customer? What implications might this have for someone without the marketing resources of a large corporation? These phenomena are broadly known as “filter bubbles.” While you wait for a checkoff, read more about them: