Social Media, Information Cascades, and Ranking
As we learned more about networks effects and information cascades, my mind was increasing drawn to the applications of these ideas in social media, and how information one sees in his/her own feed is deemed to be more valuable than the information that doesn’t appear there. Though information cascades are not a perfect description of this situation, the concept is similar. While in information cascades a person observes the decisions that others make and reasons based on that, with social media, they observe the information fed to them, and this information helps shape their worldview. That’s not to say that information cascades can’t apply directly to social media – specific aspects of social media such as showing interest in events is exactly the sort of decision-making force described by information cascades.
Perhaps a more interesting question is – how do platforms predict what content to deliver to your feed so that you will return again and again? This is very similar in nature to the PageRank exercises we did. According to the linked article, Facebook’s feed-ranking EdgeRank algorithm considers affinity, weight, and decay. Affinity, or how close the user and the content are, should certainly draw on networks. Maybe the user has numerous friends who clicked on or interacted with the content. The interesting part of the affinity metric to me is how different “types” of friends are considered differently. Facebook friends are representative of an incredibly bare-bones relationship, so the fact that many of your friends enjoy some content is not necessarily a good metric for whether it should be delivered to you. Submetrics to look at in considering “good” friends could be the amount you interact with someone’s posts, or the amount you interact with content that they also interacted with. Everything involved with calculating affinity seems to be ingrained in network theory.
Link: https://www.smartinsights.com/social-media-marketing/social-media-optimisation/social-network-algorithms/