Information Cascades in Social Media
The concept of information cascades and how easily one can be created based on very limited information is something that we have discussed extensively in lecture. We have shown how the first few people to react to something can largely determine the whole’s reaction. In the article “Information Cascades in Social Media in Response to a Crisis: a Preliminary Model and a Case Study” by Cindy Hui, Yulia Tyshchuk, William A. Wallace, Malik Magdon-Ismail, and Mark Goldberg, the authors study how information spreads after a crisis event on Twitter after a crisis event as a case study. They identify the types of information requested or shared during a crisis situation; show how messages spread among the users on Twitter including what kinds of information cascades or patterns are observed; and note what these patterns tell us about information flow and the users. For example, the article divides users in the network into categories based on how connected they are with the rest of the network and what kind of information they have received, which ultimately decides what actions they take, such as disbelieved users, who have received accurate information but are unconvinced for some reason, causing them to take no action, as compared to believed users, who have received correct information and believe it, causing them to spread the message to their neighbors. It also includes people who are disconnected from the network or people who never received the message, in which case they obviously do not take any action. One of the most interesting categories is that of the undecided user, who presumably has received a mix of accurate and inaccurate information, causing them to be unsure, so they choose to query their neighbors for their opinions. Ultimately, the study concludes that the actions of emergency managers, who are those who are tasked with disseminating actionable information, such as warnings to move to safety, have a great amount of control over the spread of information. As such, the paper suggests that they use the various insights revealed in the paper, such as when to tweet or how to tweet to maximize the spread of their information, to facilitate the spread of accurate information or impede the flow of inaccurate or improper information.
Many of the insights revealed in the paper correspond to things that we have learned in class. For example, being one of the first people to tweet generally spreads the information more than being one of the last, which makes sense because it is likely for a cascade, accurate or not, to have already been started by then. However, many of the insights go beyond what we have discussed. For example, they discovered that tweets are more likely to form long chains and webs after a “Stay in Shelter” warning, while they are more likely to be more random and dispersed after an “All Clear” message. Naturally, given the nature of the class, the ideas we are learning apply to information cascades more generally, while this study focused on a specific case, so naturally the study will create more detailed and interesting insights that the general purpose ones we are currently studying. However, it is still interesting to see the ideas we are currently learning be applied to a real life situation with potential life-or-death consequences.
https://dl.acm.org/doi/10.1145/2187980.2188173