About a week ago I saw this article on the New York Times website: “With Graph Search, Facebook Bets on More Sharing.” My first thought, of course, was “What the hell is a graph search?”
Thankfully the first day of our Networks class cleared that up for me: Graph Search. Graph Theory. Networks. Social Networks. Facebook.
This new feature (currently in beta) is Facebook’s attempt to leverage the trove of miscellaneous personal information it has amassed over the last near-decade to create a new “social”, network-based, search engine. Rather than relying solely on keywords and Boolean operators (arcane relics of the early internet) to generate results, Graph Search will utilize the likes, check-ins, and photographs of the friends in your network to present results that are personalized and, Facebook hopes, more relevant.
The premise of Graph Search is that the information available within your social network amounts to a searchable dataset of personal recommendations, and that you will find these personal recommendations more valuable than the results of a standard keyword search.
The usefulness of Graph Search’s recommendations will be largely determined by how much information users choose to make available through their privacy settings. Facebook, being Facebook, is acutely aware of network connectedness: they know that the introduction of Graph Search will affect the behavior of their users, causing these users to either share more or less information. Facebook is clearly counting on more. Tom Stocky, one of the creators of Graph Search, notes in the NY Times article that Facebook is hoping users will actively share more information in hopes of appearing “useful in the eyes of their friends.” Facebook is counting on a cascade effect in which one group of friends after another discover the social benefits of opening themselves up (even more) online, leading eventually to a network effect in which the Graph Search will find its usefulness amplified as a function of its popularity.
Interestingly, this is all happening against a backdrop of increasing concern over online privacy. The NY Times cites several studies that found young users actively reducing the amount of information they are sharing online. This makes me interested in the possibility, from a game theoretical standpoint, of cheating.
Graph Search could be understood as a public good game. It is possible for some users to cheat (withhold their information, reducing the usefulness of Graph Search for everyone, while continuing to benefit from the Graph Search feature themselves), while others cooperate (share their information, increasing the usefulness of Graph Search for all users). Without a system of punishment to discourage cheating (like tying one’s access to Graph Search to one’s privacy settings), Graph Search will not have the increased amount of data it needs to live up to its potential.
For those of you who were in Behavioral Economics with me last semester (I know I saw one or two of you), I thought I’d throw this question your way: considering that the reliance on personal recommendations introduces all manner of biases (the law of small numbers, gambler’s fallacy, and confirmatory bias), do you think Graph Search will provide results that are better or worse than what one might find using Google?
New York Times: “With Graph Search, Facebook Bets on More Sharing”
– Rob Lindgren