Information Cascades as a Result of the Structure of Social Networks
Information cascades are a complex phenomenon by which the public signals of others’ decision making has the ability to influence subsequent decisions being made by individuals even when they have their own private information. In a localized setting, information cascades can look like deciding whether or not you think you should go to a restaurant depending on how crowded it is inside. However, in the context of the internet, information can spread quickly and far, which can influence many different people’s decision-making process. As such, it’s important to analyze the features of social networks that allow information to propagate so that we are aware of the dynamics by which misinformation may spread and lead to potentially harmful information cascades.
What features of a node within a network allow it to have more control over the information that is disseminated through the network? This journal article, Social Network Analysis and Information Propagation: A Case Study Using Flickr and YouTube Networks, by Akrouf et al, helps to establish some of these factors. Akrouf and their team used visual representations of YouTube and Flickr social networks by examining the comment sections and contact books of users of each website respectively. In this case, the Flickr network is an explicit network as each link represents a definitive link between the two people communicating. The YouTube network serves as an implicit network in this case, as the network is formed by analyzing content from within the comments that link different nodes based on similarity.
What the researchers found is that the network structure is the main limiting factor in information networks. The YouTube network in particular was composed of different components with not a lot of interaction between them, which suggests a somewhat strict organization of information on YouTube. The average distance between nodes was also higher in the YouTube network than the Flickr network, which made it harder for information to traverse across the entirety of the network. Within YouTube, however, there was a lot of information dissemination within each of the giant components. It was also noted that explicit networks help information traverse easier than implicit networks, as explicit networks are more likely to create links between mutual nodes through people directly adding their friend’s friends. There were also some drawbacks to these analyses however, as not all information is spread through comments on YouTube as some may simply share videos through iMessage or other communication channels and express their views there. In addition, not all YouTube comments were helpful in disseminating information as many comments are simply spam messages. All in all, these researchers found the information dissemination process across large complex networks to be incredibly complex and not simply reducible to factors like centrality.
This research has a few key implications for how we view social networks and misinformation spread. First, it seems that the information spread within strongly connected components was more reliable than across these components, suggesting that, especially amongst YouTube social networks, there is a high degree of self-sorting “echo chambers” which can be dangerous for inducing false information cascades. With little information feeding into these components, as well as a generally wider network with farther distances between nodes, there is not a lot of rectifying information available to course correct potential cascades. However, it also seems that information can spread more contagiously across explicit networks than implicit networks, which suggests that it is more likely that it is our connected friends by which we are receiving and sharing information from, and not strangers in the comments section of a YouTube video. This may be unfortunate when considering false information cascades as this information may be coming from highly trusted sources which may be hard to convince people of otherwise. In conclusion, it is important that we consider these structural social network factors when dealing with the ramifications of information cascades in our modern society so that we can seek to limit their influence and their negative societal consequences.