i202 Fall 2010 School of Information, UC Berkeley

Marti Hearst's discussion of social search in the reading for 11/29 reminded me of Hotpot, which Google announced about a week and a half ago.  At its most basic, the feature is essentially the company's own twist on Yelp: an opportunity for people to share their opinions of restaurants and other local businesses.  One interesting departure, however, is Hotpot's use of a six star rating scale. Six stars signifies "best ever", but users are only allowed to bestow ten of these six star ratings.    

The overall concept itself is hardly revolutionary, especially in light of Places (Facebook's months-old release that has essentially adapted this idea for its smartphone-powered users).  But while Hotpot might seem superfluous, it gets interesting after considering the way Google integrates this data into PageRank search results.

Consider the possibility that you are feeling hungry for Ethiopian food, so you type the words "Ethiopian restaurant Oakland" into the Google search bar.  Right now, you'll see a series of restaurant names that populate a Google map, along with their contact information.  However, if you're an active Hotpot user who's added some friends to your contact list, the query will also return relevant recommendations (or cautionary tales) that they've posted.  Perhaps one of your 202 TAs gave Enssaro the highly coveted six stars, or maybe a classmate dished out a one star rating to all of them, angrily declaring that they don't serve "real" Ethiopian food.  Either way, your eventual decision will be affected not only by the way PageRank ranks your search results, but also by the opinions of people you know.  

While this is relevant to a lot of the NLP and vector discussions we've had over the last few classes and readings, it's especially interesting to consider Hotpot in the context of Google Goggles.  How does this modification affect the sociopolitical bias that Diaz identifies in his thesis?  Could this inadvertently lay the foundation for a kind of PageRank guerilla warfare?

What I mean is this: maybe Addis (Google's highest-ranked Ethiopian restaurant in Oakland) is not very good.  It could have risen to the top because the food editor for the New York Times is the owner's sister, or maybe the restaurant hired an SEO expert.  Maybe it used to be good, but has seen its glory days long passed.

Perhaps there is an up-and-coming Ethiopian restaurant named Selassie's Sizzle that is amazing, but the earnest owners can't get anyone in the door because they're ranked so low on Google's search results.  However, say one night a Hotpot user wanders in by accident, loves the food, and posts a glowing rating.  Others see the rating, try out the restaurant, and agree.  Bit by bit, this could precipitate an eventual jumping in the rankings (foodies post about the restaurant on their blogs, etc), and Selassie's Sizzle might eventually topple that crumbling Colossus at the top.  

PageRank already kind of does this with Yelp rankings, but I'm not entirely convinced that it really represents a "bottom-up" revolution.  Yelp is inexact and most of the reviews come from complete strangers, who may have different opinions and tastes from the user.  With Hotpot, the search engine retrieves the opinions of the user's friends, through a platform that's well integrated with the search engine itself.  In other words, it has a kind of credibility that Yelp lacks.

Power to the people--not the algorithms! :)