Blog

Evaluating Facebook Graph search as an epistemic tool

The new Facebook Graph search will allow users to conduct a much finer grained search across their networks than they currently can. It’s not hard to imagine how facebook & Bing’s relationship might be built on here for feeding into results the actions of friends on the social network (which to some extent is already implemented in Bing). As John Battelle (http://battellemedia.com/archives/2013/01/facebook-is-no-longer-flat.php) points out:

“One can imagine such functionality will create a lot of new engagement on the service. And not just from people “friending” prospective beaus or hires. Recall that when Google burst onto the scene, it prompted a dramatic response from owners of web pages, who immediately began rewiring their sites to be optimized for search. Similarly, Facebook’s Graph Search will incent Facebook users to “dress” themselves in better meta-data, so as to be properly represented in all those new structured results. People will start to update their profiles with more dates, photo tags, relationship statuses, and, and, and…you get the picture. No one wants to be left out of a consideration set, after all.”

This new tool will change the way that people think about search, and their interactions with the site. We shouldn’t just be talking about the implications as though the system won’t impact on the data, we should be considering what this functionality might look like in 5 years, if it really takes off.

From my perspective, what’s interesting about this is that facebook (and through various other projects, bing) is trying to make use of the social value of ‘friends’ and known people as informants for testimonial knowledge.  This is cool, and some of the work Merrie Ringel Morris & co are doing on “Collaborative and Social Search” plays in this area.  The principle is:

“let’s say you work at, I dunno, Google. And you want to recruit product management talent from, say, Facebook. Again, the best way to get to that talent is probably a friend. So why not do a search for “friends of friends who work at Facebook and are product managers”? Why not, indeed.” http://battellemedia.com/archives/2013/01/facebook-is-no-longer-flat.php  As John also points out, the scope for “curation via search” (my expression) through which users conduct a search, the results of which they can then post to a friend (see www.so.cl for an example tool) is pretty interesting too.

But there are some practical and theoretical risks, for example:

  • It’s not what you know, but who you know (or, more accurately it is what you know but only after the first filter – who you know – has been applied)
  • Confirmation bias in personalisation based on friendship (see as-an-epistemic-tool/” target=”_blank”>this post for commentary & link to article on this issue) – whether search is only internal, or eventually spills out onto the open web, if our results are influenced by our friendship groups (particularly in ways we might not be aware of) this raises serious concerns about the epistemic properties of the search, which we might expect to return both all relevant results (recall), and specifically results which meet the criteria we have stated (precision)
  • Exclusion of those offline, not on facebook, opting out of the system (but still on fb)
  • Gaps in data – related to the above, some people may remain in the system, but be excluded due to gaps in their own data (or indeed gaps in search capability, or end-user specificity!)
  • Failure of nuance – listing a skill often comes with nuance, or examples to contextualise the depth of the skill; this should not be stripped.
  • Privacy concerns regarding who we wish to share particular information with, this may be true in various contexts (e.g. I might not want everyone to be able to find that I’m an expert in x, particularly if the search result is stripped of context)
  • Noise – what’s to stop me from listing everything I might in order to appear in all searches (I shudder at the idea of ‘endorsements’ coming to facebook, but given the ‘like’ feature it’s not hard to imagine some sort of implicit measure here – do we want to see the day where people add “skills” (or ‘gsoh’, etc.) to facebook which their friends and family then “like” to implicitly endorse?)
  • ‘Watcher’ noise – one might be a member of a particular group because you want to ‘keep an eye’ on what’s happening, not because it’s an ‘endorsement’
  • “Old news” issues – good informants (tend) to have a quality of recency, this may be breached by some sorts of search.  Perhaps as important/more important, people may not want their old news to be ranked highly particularly in their personal life (e.g. old relationships being dredged up), or for example travel experiences (where such information is likely to be irrelevant).  It looks like this is being attenuated in the search feature, but that in itself raises issues because in some cases the age of the information will not be relevant, it is likely to depend on a context of use.
  • [EDIT: added at 16/01/2013 11:15] What is the risk of testimonial injustice (see Fricker) – the risk that some types of user’s knowledge will be marginalised (by specific agents) on the basis of their characteristics?  Is this greater or lesser than other search (and/or recommender) systems?  (This is a case of a prejudice exercised by individuals)
  • [EDIT: added at 16/01/2013 11:15] What is the risk of hermeneutical injustice - the risk that some types of user’s knowledge will be marginalised by the system, and perhaps in such a way as to make those users unaware of their own epistemic injustice?  Is this greater or lesser than in other search, social network, or recommender systems?  (This is a case of marginalisation as opposed to explicitly enacted prejudice)

I’m sure there are other issues at stake here, and I’m quite excited/interested to see how the new function plays out!  If anyone has any other ideas, let me know @sjgknight, as it develops hopefully I’ll have some more time to write something properly…now, back to those manuscript revisions ;)

[UPDATE 9:01 17/01/2013]

 


This Post Has 4 Comments

  1. Simon Knight says:

    “Even without context, Facebook is also trying to approximate real world trust. Its search engine ranks answers to every query by an awkward construct that Facebook calls “social distance.” Its algorithms vet who among a user’s Facebook friends the user is closest to and whose answers the user would like to see at the top of search results. The company is betting on the principle of homophily: if it is from someone the user likes, the user may be more likely to pay attention to it — and click on the link. ”

    http://www.nytimes.com/2013/01/29/business/how-facebook-taught-its-search-tool-to-understand-people.html?pagewanted=2&_r=2

Leave A Reply





%d bloggers like this: