Facebook announced a series of incremental updates to Graph Search yesterday -- as revealed by a peek under the hood -- and I think it’s a good thing they’re doing the feature and user release so gradually.
It might be for technology scalability and resource reasons, but it’s also good psychology. A slow and steady rollout gives Facebook time to preserve its feel of a digital living room (or neighborhood bar), a place where we can safely hang out with our friends. It also gives users time to figure out and get comfortable with where their information ends up as it is indexed and made searchable.
Because a steady supply of current preferences and intent is absolutely critical to the success of Graph Search. Yet this supply of social information on Facebook could slow as the very demand for it grows.
And it's in this tenuous balance of Facebook vs. Facebook -- not between Google and Facebook or LinkedIn and Facebook or Twitter and Facebook -- that the success of Graph Search and its impact on the company will be determined.
[#contributor: /contributors/59326d544dc9b45ccec5ddec]|||An economist, [Arun Sundararajan](http://www.stern.nyu.edu/faculty/bio/arun-sundararajan) is Associate Professor and NEC Faculty Fellow at New York University’s Stern School of Business. Sundararajan researches how information technologies transform business and society, focusing on networks, privacy, and digital institutions. Follow him on Twitter @digitalarun.|||
The Promise of Graph Search
The trouble with the dominant models of search today is that there's a tradeoff between reach and relevance.
We can exploit the reach of Google or Bing, but then we have to work hard sifting through the results to find what’s really relevant to us. Or, we can poll our social and professional networks, because homophily makes our friends reliable relevance filters, familiarity and trust legitimizes the information we get from them, and we've engineered our networks for the right mix of knowledge. But then we're limited in how efficiently we can retrieve this information when we need it.
The real promise of Graph Search is in breaking this tradeoff: It combines the relevance of social information filters with the reach of modern index-and-query search technology.
Facebook promises to put information that’s been filtered, curated, and customized by friends and colleagues at our fingertips. This information isn’t all explicitly articulated; it’s broader and deeper than the mere textual content of user profiles: It’s contextual, and it’s personal. Like knowing about the existence and quality of a restaurant because our foodie friend checked-in there, or getting a colleague’s opinion about a hotel stayed at in 2009 from a casually shared wall post, or discovering what's really fun to do in Paris as revealed through last year’s smile-filled Instagram feed.
Such knowledge is hardly stale and shallow. Add in the data trails linked to user accounts from over 40 million Facebook Connect sites, and Facebook has a veritable treasure trove.
The Perils of Graph Search
Mining and indexing its online graph allows Facebook to disrupt the current social search model, propagated from the physical world, of polling our networks for information -- through status updates like “Heading to Prague, anyone have suggestions on things to do there?”
>This isn’t a generic privacy debate. Information is the single most important economic input -- the only “raw material” -- of companies like Facebook, Google, Twitter, and Pinterest.
But here’s the rub: By moving from “poll” to “retrieve”, Facebook makes accidental or unintended information sharing far more likely.
We lose our “anonymity from obscurity” because our posts and social information are no longer for our friends or just for today -- but forever available, perhaps even to strangers, especially if we haven’t wrapped our heads around the nuances of constantly evolving privacy settings.
And while these costs and risks to users grow, the benefits from sharing information don’t really change. So the social sharing that generates the supply of information gets throttled gradually as the very demand for this information grows. Unable to assess the risks of sharing, we might simply stop doing so.
This isn’t a generic privacy debate. Information is the single most important economic input -- the only "raw material" -- of companies like Facebook, Google, Twitter, and Pinterest. A reliable and valid stream of intent and preferences is the lifeblood of such internet economy firms. It makes them die or thrive.
What’s the point of trying to “maximize searcher happiness” if it ends up minimizing the inflow of what the users came to search for in the first place?
So what should we do about this? We need to care, because we could really benefit from something like Graph Search -- its value to users is incontrovertible, and nothing else (yet) breaks the historical tradeoff of information retrieval.
Here’s one easy place to start. Facebook should allow users to retroactively choose access permission settings for every piece of their information, enabling them to specify its precise reach [Note: see correction below]. Sure, many users won’t take advantage -- it’s a cumbersome process unless automated -- but at least it signals that we, the people, are valued information suppliers and that we can more explicitly define the terms of our implicit information contract.
Without these limits, we might instead shift towards posting only content we feel is safe for a broad audience -- no different than what’s on TripAdvisor or Amazon reviews. It would just create another digital town square -- not a digital living room -- tragically devoid of those little slivers of our everyday lives that represent who we are and make Facebook so unique.
Giving users complete control over their pasts might be the only way that Facebook continues to see -- and mine -- their futures.
__Correction: __Soon after this article posted, Facebook pointed out that its "Activity Log" gives users action-by-action control over the reach of their posted content. I missed this feature, which goes a long way towards clarifying my concerns over the privacy aspects of Graph Search. However, there is still no individualized control over who can see what posts a user has "liked" or over comments the user has made to another user's post -- the audience of the post is inherited. Allowing a user to control whether a "like" is visible to just the poster or the entire audience of the original post would be a good next step, since otherwise Graph Search makes it easier for people we don't know to discover what we like and comment on.
