What if there are noises in Information Cascade?
We have been focusing on Information Cascade in class, significantly how one person’s signal/decision might potentially affect the next person and even start the cascade. However, the route of communicating the signal/decision, it’s not always accurate. Other factors might potentially influence the next person’s choice; For example, what if the signal was not accurately delivered or what if the next person was not in his/her best condition to judge from the existing information that he/she might already know. In Tho, Vijay, and Randall’s Information Cascades With Noise, the topic of how noise could possibly impact the Information Cascade further with evidence.
At the end of the chapter, we come to the conclusion that cascades can be wrong, fragile, and based on little information. This research pointed out that, with or without noise, cascades can be wrong. Noise such as observation errors does not affect the time span of the cascade as the cascade will still be able to finish within a finite time. However, the noises could delay the onset of the cascade where the research has found the error threshold that increases the delay of onset. On the other hand, the fragility of the information cascade stays the same whether there are noises in the cascade or not: a private addition signal could break the cascade in both cases.
An interesting finding that I found in the research paper was that a controlled amount of noise, which is an observation error could higher the welfare of everyone but a finite number of members in the cascade. This is against my intuition as noise has to be interrupting the communication between agents or even lead to loss of information in the way of delivering. This finding could be beneficial to platforms that can possibly weigh the trade-off between different options and decide whether information cascade might be a better option for them or not.
Source: https://ieeexplore-ieee-org.proxy.library.cornell.edu/document/7879302