Information Cascades and Ideological Views
In this academic paper, the researchers propose a model to learn ideological leaning of users by analyzing how political content follows through a network. In their model, they assume that information propagates from one user to another if both the users are interested in the topic and ideologically aligned. This relates to information cascades.
Information cascades, in the most simplest of terms, happen when people make their decisions independently of each other in a sequential order. The first person will follow their own personal signal. The second person will know the first person’s decision and therefore know their personal signal. So this means that person 2 will have person 1’s signals as well as their own. If person 2’s signal matches with person 1’s their decision will be easier. However, if they don’t match, person 2 will be indifferent between accepting or rejecting person 1’s decision. Then person 3 will make their own decision based on 3 signals. If the first two people had clashing decisions, then person 3 would be the tie-breaker and the people after would make their own decision based on the tie-breaker. On the other hand, if the signals of the first two people are the same then person 3 will make the same decision regardless of what his own signal is.
In the research paper, the researchers consider a directed social graph where nodes a is a follower of b. Because of this a views content on their timeline from b and they are influenced by b to share the content with their followers which would allow the flow of information. Sharing would include something like a retweet on Twitter or an upvote on Reddit. The research discussion how this would potentially create online echo-chambers and misinformation. I found it most interesting that the researchers were able to devise a procedure to learn the ideological embeddings that optimize an approximation of the chances of a set of information cascades.
References: https://arxiv.org/pdf/2109.13589.pdf