Information Cascades in Online Movie Ratings
As a movie lover, I usually check out the ratings for the movie that I am about to watch so that I could have a general sense. If the ratings are high, then my expectation becomes high as well. To obey the reciprocal principle, I would leave my rating and comment for each movie so that the rating website indeed evolves to a connective public good. Sometimes, I notice that my ratings have a big difference from the others and strangely enough, their comments persuade me so that I change my ratings accordingly. I realize that the information cascades also happen in the online movie ratings, in which people believe others have more information than them and hence make better decisions. Also, if some people claim that “I don’t know what the movie is talking about”, they would be perceived as inattentive and not “smart” enough to interpret trivial details.
To better understand this phenomenon, I read an article titled “Do I Follow My Friends or the Crowd? Information Cascades in Online Movie Ratings”. In the paper, researchers examine the social influence of prior ratings and, in particular, investigate any differential impact of prior ratings by strangers (“crowd”) versus friends. They employ two concepts, “Word of Mouth” (WOM) and “User-Generated Content” (UGC), to shed light on the relationship between users’ ratings of movies and social interactions. The idea of “information cascades” is covered in our lectures and textbook, meaning that “when people make decisions sequentially, with later people watching the actions of earlier people, and from these actions inferring something about what the earlier people know. Individuals in a cascade are imitating the behavior of others, but it is not mindless imitation. Rather, it is the result of drawing rational inferences from limited information”.
In the article, the authors claim that there is a “stronger association between reviews and sales in the presence of reviewer identity disclosure”. They find that informational cascades do not have a significant impact on user ratings on software adoption, but closely relate to the motivation for users to generate reviews on the Internet. The incentive to share emotions drives people to make comments recurrently with multiple people. In addition, people frequently engage in “observational learning” by making assumptions of the quality based on peer choices. Therefore, the researchers find out that people may overreact towards more positive (negative) reviews or be disgruntled by other reviews, but the communication or discussion between friends (via social networks, e.g., online friends are acquainted with each other in an online or offline community) may lead to private information flow. As a result, they propose the hypothesis that the higher the average rating by prior users is, the more likely a subsequent user is to choose a high rating versus a low rating.
They observe that herding behavior is moderated as sharing friends’ ratings may reflect the likelihood of convergence to efficient quality information of ratings. Sadly, on the other hand, the quality of reviews is strongly impacted by the presence of observational learning in online user rating because each user rating would be associated with bias from herding behavior to different extents. Drawing on informational cascades, the authors suggest ways alleviate the bias so that the integrity of quality information could increase.
First, they propose to aggregate information. Second, if users are allowed to share their product experiences as much as possible, then herding behavior could be moderated as well. Besides, as information overloads and online anonymity make information flow difficult, the researchers state that another effective solution could be increasing social interactions among online friends to flow private quality smoothly in a large-scale consumer network. Last but not least, since friend rating is less influenced in terms of observational learning in the user rating decisions, increasing the visibility of friend recommendations is another feasible way to prevent this bias.
reference: https://repository.upenn.edu/cgi/viewcontent.cgi?article=1324&context=marketing_papers