Information cascades and online movie rating
Previously, we discussed the information cascades theory and how easily the model can be formed based on very little information. It surely showed the power of herding effect – individuals converge to a uniform social behavior. (Banerjee 1992, Bikhchandani et al. 1998) However, we also mentioned that the cascades can be very fragile: if anyone can get access to better information, he/she can break the tie immediately.
So in the article, “DO I FOLLOW MY FRIENDS OR THE CROWD? INFORMATION CASCADES IN ONLINE MOVIE RATING ”(Young-Jin Lee, Kartik Hosanagar, Yong Tan, April 15, 2015) the authors explore the dynamics behind individuals’ behaviors by looking into the social network’s influence on the information cascades. They observed one phenomenon that “friends’ ratings always induce herding but its effect is weaker than that of the crowd for popular movies, and the presence of social networking reduces the likelihood of herding or differentiation on prior ratings by the crowd.” (article’s conclusion)
This research studies on the progress made on the topic of information cascade, observation learning and behavioral psychology to solve some difficult challenges and effectively show the readers how complicated the whole system can be. There is a scenario which is very relevant to the real life situation: if we know a friend is often critical in her movie assessment, then this friend’s low rating may be less salient and influential. However, if this same friend gives a very high rating on a movie, the compliment definitely values more than other strangers’ comments.
Furthermore, they study the question “How does social influence relate to the level of social interaction among users?” They demonstrate the way to restructure the cascade theory in this context to understand how prior rating by the crowd and friends may influence the following users differently:
- The most essential principle: people observe others’ actions and make the same choice independent of their own private signals(Banerjee 1992, Bikhchandani et al. 1992, Banerjee 1993, Bikhchandani et al. 1998)
- Every user’s private information is his/her own after watching the movies. All movie related information can be publicly viewable.
- Here comes the interesting part, even if every user can get access to other’s ratings, since all individual reviews are often infeasible, hence people may only view aggregated information such as the average rating for the movie (Li and Hitt 2008).
- Since the socially closer relationship, people have better context to understand friends’ reviews and ratings.
But there are various factors that can influence the shape of the system, so various challenges arise. The authors eliminate other sources of social influence such as network externalities or word-of-mouth communication and consumer heterogeneity by creating an effective and innovative dataset supported by rich and reliable information. And finally they successfully conclude that “The herding effect for popular movies can lead a user to provide a more positive (negative) rating in response to higher (lower) prior ratings by the crowd, whereas a differentiation effect for non-popular movies can lead subsequent ratings in the opposite direction. In contrast to crowd ratings, we find that only a herding effect is associated with friends’ prior ratings, regardless of the popularity of a movie.”
One of my biggest takeaways is to learn how to evolve the basic information cascade model/ one quintessential principle in a way to be applicable with more complicated scenarios. And also consider the possible deficiencies along the evolution of the model so that the future research can gather new information to study on the topic in a more comprehensive and well-grounded way. Also, the topic examines the impact of information cascade under the perspective of degree of social connections, which is very interesting, meaningful and innovative.
https://pubsonline.informs.org/doi/abs/10.1287/mnsc.2014.2082