Social influence of hubs in information cascade processes
The “social influence” factor is important to consider in the information cascade process. In class, we discussed how an information cascade is an example of sequential decision making in which people decide to either use or reject their private information when they are privy to the decisions of others. However, our models in class did not consider the relative social power of the individuals that were making decisions. For example, in the colored balls in an urn scenario, the third person that announces their choice is weighing person 1 and person 2 equally. This would not be the case if person 3 has reason to believe that either person 1 or 2 is more or less reliable/has more social influence over the subsequent individuals. This idea is the premise of the article “Social influence of hubs in information cascade processes” which provides a computational simulation approach to cascades, specifically how cascades can be misled by incorrect information. Implications are relevant in the field of marketing, sequential voting, and technology/innovation adoption.
The article specifically mentions the PageRank algorithm, and how hubs can make incorrect decisions that can cause a misled shift cascade. As discussed in class, PageRank updates the nodes at each step based on the sum of all the shares it receives from its child nodes. The article also mentions eigenvector and betweenness centrality as correlated with information cascades among agents. Social influence is relevant in this case as evidenced by an econometric analysis of how hubs incorrect decisions can lead to market failures.
The idea of social influence is particularly important during elections, particularly the U.S presidential election which can be seen in the cascade that forms after the Iowa caucus and Super Tuesday. This can be modelled as an information cascade because people may disregard their private information (who they personally believe the best candidate is) in favor of voting “with the crowd,” even if the voting with the crowd is against their personal beliefs. The Iowa caucus and Super Tuesday are essential for this precise reason – success in early primaries often predict campaign success later on (unless the cascade is broken later on as seen with Joe Biden in 2020.) Information cascades may also contribute to why so many Americans feel that their vote doesn’t matter, and thus why some do not vote, especially in states with few electoral college votes. Once a certain candidate has gained enough traction, at a certain threshold that support will be sufficient to win them the election because their support will eventually influence enough other people. This relates to social influences where states with higher electoral votes have more social influence and their votes are therefore valued more significantly than the states following. Once electoral votes have gone to a certain candidate, it no longer matters who other people want to vote for because the threshold has already been met. So, are information cascades ruining our democracy? Probably not, but it is important to recognize the helplessness that people feel during elections to vote the same as others.