Facebook’s graph search is going live for hundreds of millions of English-speaking users today, the first serious option for searching a social-media site that blocks Google’s access to much of its data. Facebook claims its first foray into a powerful search tool will be a boon for its users, who will then stay on the website longer—and that, in turn, will be a boon to its advertisers. Everybody wins.
But like many tech products, Facebook’s new search tool is being launched in an awkward early form of its development. The company hopes we’ll stick around as it improves, and that could very likely happen. But there are several open questions about graph search.
Can it learn to understand us?
Part of the alleged joy of Facebook’s graph search is the way users can just type in a question as they’d ask a person for the same information. A truly conversational search would give it a leg up on Google, whose basic search tool is based on keywords. But this trick, known as natural-language processing, works only if it really works, and early reports have found Facebook flummoxed by wrinkles in language that even the densest human would have no trouble understanding. Facebook’s graph search apparently doesn’t realize that “surfers,” “people who like to surf,” and “people who like surfing” are the same group of people.
Of course, systems like this are often bad at first; they improve by gathering data. That’s why Apple released Siri when it wasn’t quite ready for prime time. On the other hand, is Siri much better now? I haven’t noticed. For now, Facebook’s task should be easier than Apple’s. The data it is accessing are Facebook’s own, which have been pretagged with categories that Facebook itself created. It would be harder if it had to wade out into the open Web, where things are weird and hard to understand.
Is it pulling the most relevant information?
For many Facebook users, the activities that keep them coming back to the site are checking and updating the news feed. These status updates contain lots of juicy information and can be a pain to search once they’ve scrolled past you. But the early version of graph search will not include information from status updates—meaning graph search misses a lot of relevant information. It’s also left with some of the less valuable material to pore over. Graph search will rely largely on what people have said they like. Of course, everyone knows there’s a gap between what your “Facebook self” likes within the context of the social network and what you actually like as a person. You press that button on the site for any number of reasons.
Can it go beyond Facebook?
If graph search is ever going to be important, it will have to leave Mark Zuckerberg’s walled garden. The social network’s strength is its breadth—who doesn’t have a Facebook page? But increasingly, depth is the problem. Many of the most interesting questions can be answered only by accessing data collected on someone else’s server. Do my friends in Cleveland like Bone Thugs-n-Harmony? Spotify probably knows better than Facebook. What is my brother’s favorite takeout restaurant? Seamless could be the place for that. What are my college friends’ favorite books? Hello, Goodreads. And of course, there are questions that Google will likely always win on: How tall is Kareem Abdul-Jabbar? Some data repositories, such as Facebook’s own Instagram, will be easy. But good luck with data held by the company’s major competitors.
Will it creep everyone out?
Companies aren’t the only potential participants likely to be skeptical. Every new service Facebook rolls out comes with its own privacy scare. Since graph search is built to help people pluck meaning from the overwhelming amount of data on Facebook, it will necessarily complicate the anonymity-by-obscurity strategy favored by some users. Facebook is also planning to use the product as an enticement for advertisers. Like most of Facebook’s tools, the potential threats can be neutralized by savvy navigators of privacy settings. That won’t keep the company from coming under criticism, of course. But given recent history, vague (or even very, very specific) privacy concerns aren’t enough to drive people away from free services online.