Information Cascades on Yelp
http://snap.stanford.edu/class/cs224w-2011/proj/emmao_Finalwriteup_v1.pdf
With the power of the internet today, people can acquire information in a matter of seconds. On the other hand, people can also provide information and share their opinions on various platforms and topics. One of the most common uses of Google and other search engines is looking for food, transportation, and activities. Specifically, platforms like Yelp facilitate the transfer of information through restaurant and establish reviews. When people assess what restaurants to try, they often look towards the reviews of people that have been there on Yelp.
Since Yelp is a platform where information is transferred from one person to another, it is interesting to speculate on how ideas like information cascading can play a role in how users pick restaurants. Information cascades occur when users make a decision based on the decisions of people that participate in the activity before them, rather than the actual value they get from an activity. In the Stanford research study called “Identifying Cascades in Yelp Reviews,” Grace Gee, Chris Lengerich, and Emma O’Neil examine the trends of reviews for particular restaurants with many reviews. Particularly, they found that many restaurants’ reviews follow a cascade rather than random reviews, but overall cascades are not dominant.
To test the idea of cascades in reviews, the researchers focused on distinguishing good and bad reviews. They then examined the reviews in time order, to see if there were cascades of good reviews or bad reviews. They ended up finding that for some restaurants, good cascades can be quite prevalent. Specifically, when people read good reviews of a restaurant, they are likely to have a more positive pre-emptive perception of the restaurant and tend to maintain this image. This makes sense that if people see that a restaurant is regarded highly, they are more likely to enjoy it, but these are not true cascades because people are still taking into consideration their own ‘values’ and experiences in this situation.
Interestingly, the researchers found that bad cascades were incredibly rare on yelp. They found that there were almost no long consecutive streaks of only bad reviews. Perhaps, this follows the cascade model, as people are less likely to go to a restaurant if they see that the last review was negative. Thus, a bad cascade cannot form on Yelp, since people will not add to the cascade if the person before them gave a negative answer.
Overall, it is incredibly thought-provoking to consider the implications of information cascades, especially through online platforms like Yelp. People online are so impressionable, and we often make many decisions in our lives based on what Google tells us. Thus, it makes sense that good cascades can form for Yelp reviews, but bad cascades do not really occur because people will not continue to go to the restaurant if the person before them gives a negative review.