Social Search Redux


A week or so ago, we wrote a post on Social Search, and how (we believe) it is different from Collaborative Search.  We have also begun laying out a taxonomy of the various factors or dimensions that characterize information seeking behaviors involving more than one person.  So far, we have listed two dimensions: Intent and Synchronization.  We will continue with two additional dimensions over the next few weeks: Depth and Location.

But in the meantime, we note that Intent and Synchronization already give us enough material to draw descriptive and discriminatory lines between various types of multi-user search.

Multi-user search can be:

  1. Explicit  and Synchronized
  2. Explicit  and Non-Synchronized
  3. Implicit and Synchronized
  4. Implicit  and Non-synchronized

Most of our current research at FXPAL involves category (1): explicitly  collaborating searchers (explicitly shared information need) and whose search activities influence each other bi-directionally (synchronized).

On the other hand, most of the research involving search by multiple people  involves category (4):  implicit collaboration based on aggregated actions of past searchers that later influence a single searcher’s activities.  In this case, the influence does not flows only in one direction.  Examples of approaches that belong in Category (4) are collaborative filtering and recommendation systems.

With this taxonomy in mind, we can return to our discussion of a term that has become popular in recent months and years: “Social Search”.  In our understanding, Social Search belongs in (is a continuation of) Category (4).  In social search, people don’t seem to be actively dividing up a task and sharing responsibilities for finding information jointly relevant to a group.  Rather, social search (as far as we’ve seen it discussed online and in the literature) is about non-synchronously using information that your friends have found to bias your search results toward information that they have already found useful. Social search is about re-covery and repetition (i.e. propagation) of already-found information, rather than the dis-covery of information new to the group as a whole.

An article by Ramana Rao at the CACM blog drives this point home.  In this article, the author uses light as an analogy for information flow in social search.

But, first, let’s consider people.  The notion of wavicles, which captures the simultaneous particle- and wave-like properties of light, fits somewhat to our social existence.  We are individual particles and participants in wave-like social phenomena that transcend us individually.  With this bi-focal lens, you can see how social filtering operates at the two levels and its somewhat-ness as I noted:

Here, Rao seems to be confirming the non-synchronization of Social Search.  The actions of the crowd hit you like a wave, and your own search actions flow back as a particle.  In social search, you are not directly sharing a search task with other people: influence is non-symmetric and non-synchronized. Moreover, the particles that you create do not necessarily return directly to the same people whose aggregated wave hit you in the first place — the collaboration is implicit.

To be sure, there are lots of open and interesting research problems in the Social Search realm.  But collaborative search, as characterized by explicit sharing of information needs and synchronized (bi-directional) influence during the fulfillment of those information needs, represents a fundamentally new and different way of approaching multi-user search.  Collaborative Search is not Social Search.


  1. Am glad that youre making this distinction between collaborative search and social search. I believe this is a good step forward to differentiating the various research efforts out there.

  2. Oh, folks reading this might be interested in trying out, which is a social search engine built on top of social bookmarks. Based on other peoples bookmarking patterns, it recommends URLs for the search queries youre interested in. The underlying algorithm uses a bayesian network modeling approach. This would satisfy what is called social search here.

  3. jeremy says:

    Thanks Ed! I saw you give the mrtaggy demo at PARC a year ago or so. I quite liked it!

  4. […] be an example of social search, but I recognize that I use the term in a broad sense. Perhaps, as Jeremy suggests, it#8217;s better to think of social search and collaborative search being different aspects of […]

  5. Twitter Comment

    More info on collaborative search mentioned by Marti Hearst at #ESS10 [link to post]

    Posted using Chat Catcher

  6. Twitter Comment

    RT @HCIR_GeneG: More info on collaborative search mentioned by Marti Hearst at #ESS10 [link to post]

    Posted using Chat Catcher

  7. […] Tunkelang on his blog. While she did touch on topics covered in her book, including some of the collaborative search work done here at FXPAL, she has shifted her focus somewhat to address the more social issues […]

Comments are closed.