Skip to main content



The Subjectivity of User Reviews

In the information cascade model presented in class, the world is assumed to be in one of two possible states for everyone. Either it is a good idea to make a particular choice or a bad idea. The herding experiment and its Bayes’ Rule model are explained using the urn example where the urn is either majority blue or majority red. In this case, there is a constant, correct option. However, in real life, often there is no unanimously correct option. Not all reviews are positive even for the most popular restaurants. For instance, Ithaca’s Moosewood Restaurant is an award winning restaurant. Yet, people who are not accustomed to vegetarian meals are not likely to give high reviews to Moosewood. Two people may give opposing reviews to Moosewood and both of them could be true.

Moreover, with websites like Yelp, the decision making process has changed greatly. One doesn’t have to make decisions based on the number of customers at a restaurant. In addition to their own private information, every person also has access to other reviewer’s private information online. On Yelp, every restaurant gets a numerical rating out of five and detailed reviews by individuals. The numerical rating does not reflect the subjective nature of reviews discussed above. If a person X is reading reviews for a particular restaurant, it’s important to check X’s predilection for that cuisine. Unless the reader and the reviewer have similar views on the cuisine itself, the review is irrelevant to the reader.

A comparable website is ratemyprofessor.com where students rate professors on clarity, helpfulness and easiness. Every reviewer also shares the grade they received in the class they took with the professor. A student who received a bad grade maybe more likely to give a poor review and a student who received a good grade may give a better review. The grade can serve as an indicator for the review’s relevance to the reader. For websites like Yelp and RateMyProfessors, an easy way to connect users to relevant reviews is to find reviewers are most similar to the reader.

As the amount of information increases and becomes more easily accessible online, information network models need to evolve to reflect and include these changes in order to make the best of it.

Websites talked about –
http://www.yelp.com
http://www.ratemyprofessors.com

 

Comments

Leave a Reply

Blogging Calendar

November 2014
M T W T F S S
« Oct   Dec »
 12
3456789
10111213141516
17181920212223
24252627282930

Archives