Search Takes a Social Turn

Now, even on the Internet, it is not what you know but who you know.

Michelle V. Agins/The New York Times

Chris Dixon, left, and Matt Gattis of Hunch, a start-up in New York that tries to predict things its users might like. It uses data on topics and articles from users’ Facebook and Twitter accounts.

After a decade when search engines ruled supreme — tapping billions of Web pages to answer every conceivable query — many people now prefer getting their online information the old-fashioned way: by yakking across the fence.



Turning to friends is the new rage in the Web world, extending far beyond established social networking sites and setting off a rush among Web companies looking for ways to help people capitalize on the wisdom of their social circles — and to make some money in the process.



“What your friends think and what people like you think is much more relevant than what everybody thinks,” said Augie Ray, an analyst with Forrester Research.



Amazon.com now allows its shoppers to connect to their Facebook accounts so that Amazon can display their friends’ favorite books, films and other products. TunerFish, a start-up owned by Comcast, lets users share what television shows and movies they are watching, mapping out an up-to-the-minute TV guide of programs gaining in popularity among their friends.



And Loopt, a location-focused social network with 3.4 million registered users, recently began showing them which of their friends liked a particular restaurant.



“We’ve gotten a tremendous response from that,” said Sam Altman, a co-founder. Mr. Altman said that one’s network of friends “is an incredible predictor of what you will like.”



On Google and other search engines, searches for things like hotels or electronics can turn up a lot of online clutter and spam. Instead, many people informally poll their friends for recommendations, often through social networks like Facebook and Twitter.



“Improving search has always been about improving relevance,” Mr. Ray of Forrester said. “But the thinking now is that getting information from your immediate social network is what will really make results more relevant.”



While user-contributed review sites like Yelp and TripAdvisor have long been popular ways to get a quick reading on a new place to eat, the sheer volume of reviews they offer can be overwhelming.



And the reliability of those reviews can be hard to gauge. Some may have been planted by management, while others are from disgruntled customers with a bone to pick.



The trust factor of friends’ suggestions can make a big difference. Mr. Altman said Loopt’s users are 20 times more likely to click on a place their friends had liked or visited than a place that simply ranked higher in search results.



So-called recommendation engines on sites like Amazon and Netflix try to guess what customers might like by comparing their previous purchases or rentals with those of others with similar tastes. But that approach often does not offer much insight as to why a particular film or restaurant is being recommended, said John Riedl, a professor of computer science at the University of Minnesota.



Social networks, he said, “do a richer job of constructing recommendations.” For example, seeing that a friend is frequenting a new pizzeria can have a lot of influence over whether you go.



Of course, your friends are not generating the amount of data that a company like Amazon may use to make its automated recommendations, which could result in fewer choices. “The trade-off is that you will be more comfortable with the recommendation,” Mr. Riedl said.



TV watching, often a solitary activity, is an obvious candidate for some social tips. TunerFish shows which programs are gaining in popularity in your online social circle, and what is being watched right now.



Although TunerFish is available only on the Web for now, the company says it could eventually be brought to the TV screen through an application running on a set-top box.



Facebook has its own recommendation system in place. The service allows its 500 million members to click a button to indicate what news articles, companies and celebrities they “like,” and it shares data about those preferences with its Web partners. When a Facebook user visits a Web site like Yelp or TripAdvisor, they are shown reviews from friends before they get to those from strangers.



Facebook recently began introducing a feature called Facebook Questions that allows users to pose questions to friends and strangers using the site. Last month it introduced a service called Places that encourages people to “check in” at places they visit and broadcast their location to friends.



The new services will help Facebook amass even more data on its users’ tastes. But for now there is no comprehensive way to search through or refer back to the information your friends have shared.



Bret Taylor, the chief technology officer at Facebook, said the company’s main focus was on helping other sites add social features. But he said the company was thinking about ways to corral the “likes” and suggestions of its members into a more cohesive system.



“Exposing higher-quality recommendations in more obvious and prominent ways would improve the health of the system,” he said.



Hunch, a start-up based in New York, wants to go beyond cataloging the places and products for which your friends have already expressed affection. With some complex software, it tries to use that information to predict what other things you might like, even if nobody you know has ever offered an opinion on those things before.



The service pulls in data about articles, topics and people that you and your friends have “liked” on Facebook or follow on Twitter. “Based on your placement in the social graph and who your friends are, we can make inferences about what you like,” said Chris Dixon, who founded the company with Caterina Fake, one of the creators of Flickr.



For example, Hunch’s database shows that people who are big fans of Twitter are also likely to be interested in visiting the Museum of Modern Art in New York, while the non-Twitterati tend to favor the theater.



Mr. Dixon said the company was testing a local search tool that can make restaurant, shopping and hotel recommendations. “We can take your taste profile and data from the Web and begin to match you to places that you will like,” he said.



The company says it plans to unveil partnerships with major e-commerce, news and travel sites, along with mobile location-based services.



A shopping or travel site that Hunch is working with could help a visitor decide which offerings would best suit their tastes. Based on the data Hunch has collected about correlations between user preferences, the site might suggest that someone in search of a hotel in Las Vegas should stay at the Venetian if her online circle of friends listens to hip-hop artists like Rihanna and Usher.



The friend trend, where likes matter more than links, could eventually present a significant challenge to Google, which has struggled to create appealing social services.



In February the company introduced Buzz, which lets Gmail users share updates and photos, and it is including those updates in its customized results when users perform Web searches.



“People are likely to find what your friends are saying about the iPhone 4 or a Chinese restaurant more helpful in a Web search,” said Matt Cutts, a software engineer who oversees search quality at Google.



Mr. Cutts declined to talk about what Google might do next in the social search area, but he did say he expected “these sorts of trends to continue.”

SOURCE : NY Times