Skip to Content

People are flocking to online social networks. Facebook, for example, claims an average of 250,000 new registrations per day. But companies are still hunting for ways to make these networks more useful–and profitable. In the past year, Facebook has introduced new services aimed at taking advantage of users’ online contacts (see “Building onto Facebook’s Platform”), and Yahoo announced plans for an e-mail service that shares data with social-networking sites. (See “Yahoo’s Plan for a Smarter In-Box.”) Now a company called Delver, which presented at Demo earlier this week, is working on a search engine that uses social-network data to return personalized results from the larger Web.

Liad Agmon, CEO of Delver, says that the site connects information about a user’s social network with Web search results, “so you are searching the Web through the prism of your social graph.” He explains that a person begins a search at Delver by typing in her name. Delver then crawls social-networking websites for widely available data about the user–such as a public LinkedIn profile–and builds a network of associated institutions and individuals based on that information. When the user enters a search query, results related to, produced by, or tagged by members of her social network are given priority. Lower down are results from people implicitly connected to the user, such as those relating to friends of friends, or people who attended the same college as the user. Finally, there may be some general results from the Web at the bottom. The consequence, says Agmon, is that each user gets a different set of results from a given query, and a set quite different from those delivered by Google.

“We have no intention of competing with the Googles of the world, because Google is doing a very good job of indexing the Web and bringing you the Wikipedia page of every search query you’re looking for,” says Agmon. He says that Delver will free general search queries such as “New York” or “screensaver” from the heavy search-engine optimization that tends to make those kinds of queries return generic, ad-heavy results on Google. “[As a user], you’re always thinking, how can I trick Google into bringing me the real results rather than the commercial results?” Agmon says. “With this engine, we don’t need to trick it at all. You can go back to these very naive and simple queries because the results come from your network. Your network is not trying to optimize results; they just publish or bookmark pages which they find interesting.” As a consequence, the results lean toward user-generated content and items tagged through sites such as del.icio.us.

A person can improve the results he gets from Delver by registering and allowing it access to connections made on sites where information is usually kept private, Agmon says. Registered users can also add connections found through Delver, such as a friend of a friend who consistently leads to interesting sites. Although the registration feature will be available, Agmon says that it’s important that people be able to use the site without registering, so it will be available to more casual Web users. One consequence of this design as it currently stands is that it’s possible to search the Web as someone other than yourself. Agmon acknowledges the possibility and its potential for use and abuse, but he notes that once a person builds a profile, he must log in to search, and that identity can no longer be used as a proxy.


Help from friends: Before querying Delver (above), a user enters her name to allow the site to search for information about her social network and connections. Once the network is built, the user can search and receive results ranked by ties to her social network. The query “New York,” for example, is more likely to return a friend’s home video uploaded to YouTube than to return the home page of a five-star hotel.

Clay Shirky, a professor in the Interactive Telecommunications Program at New York University, points out that Google itself is already a form of social search, due to its dependence on links. “I think the question is, what value do you get out of being more explicitly social?” he says. Since personalized results are one of Delver’s main selling points, Shirky says, it’s important to be aware of when a user wants personalized results, and when that might not be desirable. “The further we are away from our own areas of expertise and concern, I think the less we care that our results be different,” he says. “If I want to know what happened in the Crimean War, and you want to know what happened in the Crimean War, we’re pretty happy that the Wikipedia link is the number-one link, because we’ll get a good overview. If you’re a Crimean War scholar, on the other hand, you’re probably not happy about that.”

Shirky says that he is suspicious of the claim that social search sites can return better results for general search queries. “People’s tastes are more different than the same in environments with large amounts of freedom,” he says. He worries that social search results to a general query like “music” might return either a cacophony of songs that would be all over the map in terms of style, or a list of the top 40 songs no one needs to search to find. A better situation might arise for someone looking for “bluegrass music,” Shirky says, provided the user is someone, such as a bluegrass musician, who is likely to have a network well versed in the subject.

Agmon says that Delver plans to make money by serving up personalized advertising along with its search results. The company also hopes to get income by licensing its search technology to socially oriented sites and by bringing up more meaningful search results within those sites. Delver is slated to launch next month, after its crawlers have had more time to collect data. Agmon says that he expects to open the site for an invitation-only preview by May.

Keep Reading

Most Popular

Large language models can do jaw-dropping things. But nobody knows exactly why.

And that's a problem. Figuring it out is one of the biggest scientific puzzles of our time and a crucial step towards controlling more powerful future models.

OpenAI teases an amazing new generative video model called Sora

The firm is sharing Sora with a small group of safety testers but the rest of us will have to wait to learn more.

Google’s Gemini is now in everything. Here’s how you can try it out.

Gmail, Docs, and more will now come with Gemini baked in. But Europeans will have to wait before they can download the app.

This baby with a head camera helped teach an AI how kids learn language

A neural network trained on the experiences of a single young child managed to learn one of the core components of language: how to match words to the objects they represent.

Stay connected

Illustration by Rose Wong

Get the latest updates from
MIT Technology Review

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

Explore more newsletters

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at customer-service@technologyreview.com with a list of newsletters you’d like to receive.