Will The Social Graph Change Search?- All eyes seem to be on the growth of Facebook and how it is using the "social graph" -- i.e., the social data showing who is connected with whom -- to build a business. Can social data be applied to improve search? If so, would it shake up the current search space? And what impact might it have on SEO? This session explores the hype, reality and possibilities. Moderator: Chris Sherman, Executive Editor, Search Engine Land Q&A Moderator: Danny Sullivan, Editor-in-Chief, Search Engine Land
Speakers: Aditya Agarwal, Director of Engineering, Facebook Sean Lyndersay, Principal Program Manager, Live Search, Microsoft Cris Pierry, Senior Director of Product Management, Yahoo! Search
Chris Sherman is up first. What's social search? Simple definition - Internet wayfinding tools informed by human judgment. Informed can mean many things, including egregiously uninformed. There really is no "standard" definition of social search.
What's the social graph? A picture of the search behavior of a group of people, making "unseen" connections between them. It's most effective when the group is made up of trusted people.
The very first guide to the web (Tim Berners Lee, 1990) was somewhat social. We've always had social search. Tim wrote down a bunch of sites he knew of and descriptions. Then Yahoo was originally created by a team of human editors. Mta tags were created in 1996 to help content owners influence search engines and were a massive failure
Algorithmic search is social - fundamentally, search engines reflect human bias (programmer choices). Also, search engines observe human behavior - click paths, popular URLs, etc and use this to modify algorithms (Yahoo processes 14TB of user behavior a day). New personalization efforts are also used to refine search for everyone.
Why, then, did social search emerge now? Algoirthmic search has plateaued and leveled out. Nothing is dramatically innovating. Innovation is harder than it used to be. The major players have a user base that is very well established and they can't afford to risk changing anything because they may alienate their users. Humans are still better at some things than computers. (ex: Flickr) A major factor: many if not most of the players in social search are leveraging the work of volunteers. People are donating their effort to organize things and to improve services. That's a huge competitive advantage. You don't need a paid work force.
Types of social search: - Shared bookmarks and web pages: relying on people to find interesting content and share it - delicious, MyWeb, Shadows, Furl, Diigo. They are social but they are, for most users, used as a personal tool. - Tag engines (blogs and RSS): also called "taggregators" - primarily searches blogs and RSS feeds: Technorati, Bloglines, Ask Blog search, Blogpulse. - collaborative directories: created by teams of volunteers - Open Directory Project, Prefound, StumbleUpon, Mahalo, Wikisearch and Wikipedia - Personalized verticals: search service that focuses on a narrow slice of the web. It used to be difficult or labor intensive to create a specialized search engine or directory, but not anymore. Google Custom Search Engine, Eurekster's Swickis, and Rollyo are examples. - Social Q&A sites: They've been around forever - Aska's, Usenets, BBSs in the past. Yahoo Answrs, Answerbag, Allexperts today. - Collaborative Harvesters: users nominate interesting content and others vote for it. Examples: Digg, Newsvine, Reddit, Sphinn. Aggregators include Original Signal and Popurls.com
What are the problems? There's a herd mentality that dominates collaborative harvesters. Friends support friends. DIgg has a stealthy "bury" squad that it refuses to acknowledge. You also need a catchy headline to get the attention of others and it trounces substantive content.
Part search, definitely social: Facebook, Craigslist, Judy's Book, Insider Pages, LinkedIn, Squidoo
Other problems: * Scale and scope - too much expansion of the web, particularly now that there's video. * We also have issues with tagging - language is ambiguous. Orange - a juice, a fruit, a color? There's a lack of controlled vocabulary. There's also human laziness and idiots who don't know how to tag at all. * Spammers: it's a bad thing and there are no algorithmic safeguards. Users get frustrated because spammers seize control.
What will ultimately work: - A combination of algorithmic and people-mediated search (Ask's new Edison Algorithm - anytime somebody clicks on a search results, they take a title that the person clicked on and use those as tags for the page in a library. This tag library is an amazing approach. Yahoo is doing the same thing with Yahoo Answers) - Trust networks: you trust people to show you content. - Increased personalization and user control over result filtering - Social search will probably work best for non-text content (photos, music, video, widgets, etc). Why? Search engines are still fundamentally searching text but they're not so good with non-text content.
Social search is disruptive: - Social search is already impacting algorithmic results and will have an even more significant effect in the future
Social search will become more popular and important over time. People are less predictable than algorithms so you should expect potential problems. Take advantage of the tools.
Aditya Agarwal mentions that he believes that Facebook is the map of social connections. Sean Lyndersay from Live Search says that there is a problem with trust. Over time, search engines will be exploring this space and figuring it out.
Q: Why are you investing in it? AA: The type of content being generated is always changing and is always being created. Social media allows us to create accurate mapping of the social graph. Any data that helps us do this is helpful. SL: We need to understand the difference between the edges of the social graph. There's an explicit social graph - friends and family who have shared experiences with me. Outside of that are implicit social graphs: based on watching my behavior, we have commonalities (but we don't know each other). When it comes to the latter half, for search engines, in many common scenarios, recommendation is a huge part in that system. Social media is often where people talk about other people. It helps people make informed (and uninformed) decisions about the products that they buy and the services that they used. Search engines have a lot of work to do to figure out what's good and what's bad. The social media -- people's opinions -- are very critical of being able to complete the task or action that they chose to do. Social media has an edge over generic media with respect to how people find things.
Q: A lot of search marketers reach out to social media communities and there's backlash. Do you have suggestions on how to proceed? SL: This is obvious but authenticity is the key. If you're trying to pull one over the community, you're wasting your time. The stuff that attracts people's attention is humor, coolness, and good content. Do I have any tips? Not so much. It's what you shouldn't do. Don't go to Digg and plug yourself. Don't create a fake blog. You should make it easy for people to recommend your content, though. Add those social media buttons. AA: The most interesting thing is to write creative and engaging content. What's construed as marketing and what is construed as useful to users? Think about the latter. Put another way, don't try to game the system.
Q: In the peer search space, we have an outbreak of privacy concerns. In social media, a lot of the users want to put as much information about themselves as possible on the websites. What steps can they take to protect themselves? AA: Facebook helps you communicate: share information with friends. We try to give the users to control how and what they share. Privacy is hugely important to us. SL: A lot of people don't expect that this is happening when they put things in the system. Search engines are just crawling information but then people realize that the stuff that they entered 4 years ago now show up on the top of the page. We want to make sure that user know what the intention is and nothing beyond that. AA: There's a delicate balance between using the information and making sure that the information is not being used for bad purposes.
