Today I gave a talk at the Computer Supported Cooperative Work (CSCW) conference in San Diego in the Social Sensemaking track. It was called: “Towards a Model of Understanding Social Search” (in collaboration with Ed Chi).
My slides are posted here and I’d welcome any further discussion, commentary, or questions on the talk or of “social search” in general. Thanks!
4 Comments
Hi Brynn,
Came across your paper at SSM2008 and also the slides here. I find it interesting that you define social search as the social interactions surrounding the search rather than the search itself, which is where most social search players (including the one I work in) would put it. That’s surely food for thought. Two questions here:
1. Transactional queries also include goals such as buying online, reserving vacations etc. I’d expect such queries to benefit immensely from social inputs, especially being a filter against the large volume of commercial search results. And yet you said “it is unlikely
that socially-augmented search would improve or
facilitate transactional or navigational information retrieval” – is it because there were no such specific instances in the survey?
2. All major search engines today use query and click logs to suggest related queries to users. That’s in line with what you suggested for “during search”, but I can’t see how that has the same value as your before/after search suggestions, as it does not have the strong qualification from “people I know”. Did I miss something there?
Cheers,
Ofer
Thanks for your comments, Ofer. I do believe that search is an extended process, deeply embedded in everyday actions and activities—it’s therefore appropriate to study the social interactions surrounding search. In any case, current web tools don’t have great support for user-user interactions in the *act* of searching yet.
In response to your questions:
1) You’re right that an online “transaction” includes buying online, reserving vacations, etc. The web search behaviors identified in our data were borrowed from Broder’s classifications, where transactions are: “where I can perform a certain transaction, e.g. shop, download a file, or find a map”. Transactional searches were relatively rare in our data set, and we didn’t see much shopping at all. I can imagine that studying “live” search acts may reveal more instances of such transactions, so your point is well taken that social interactions may still be useful in these situations.
2) Another good point! Some sites are beginning to suggest related queries/keywords to searchers, but I agree that this won’t have the power of recommendations from your social network. (Keep in mind that our design suggestions were based primarily on the data we collected.) Take a look at Friendfeed’s model of search—when you search for something using their search tool, the result page first displays matching content that’s been shared from your Friendfeed social network. So you could (in theory) improve the “during search” phase if you can provide suggestions based on what your network has done/shared.
Thanks Brynn. I wholeheartedly agree that search involves social interactions, but I also think that to involve them directly in the search act itself is inherently difficult. There is a built-in contradiction between achieving high recall (or coverage) in results, which requires a target database as large as possible (and not only the friends that are physically near you and/or available online), and achieving high “social” precision, which demands that you concentrate on whatever social hints you have.
If you look, for example, at Mechanical Zoo’s Aardvark (aardvark.im), it focuses on precision only, producing results solely from the available social circle. Try to answer a classic informational query there, and you’d probably end up going back to Google. And at Google you have the other extreme, relying only on authority and mass coverage. Finding a golden path between these two is pretty much the holy grail of social search.
As for point #2 – absolutely, and that’s exactly what we (delver.com) do, though without the focus on real-time that FF has. But I’m not here to advertise, was just trying to understand if by any chance you stumbled upon that holy grail
Thanks again and good luck,
Ofer
Ofer, it would be great to talk more offline (although it’s probably useful for the community to make use of our discussion, recorded here).
I didn’t realize that you worked for Delver, and yes, Delver does the search-in-social-networks part fairly well. I absolutely would include your model of social search to be up there with FriendFeed’s as one to look towards. (Also, I’m not convinced the real-time component of FF is best. Maybe some combination of real-time plus popular or common actions by your friends?)
I just wrote about my experience using Delver tonight—perhaps you’d be interested in my results!
One Trackback
[...] and concluded with some practical suggestions. After watching the presentation slides, I also exchanged some thoughts with one of the authors, Brynn [...]