Facebook’s polluted graph


Cute kids. Cute puppy. Cute extortion scheme.

The Heisenberg principle states: the act of observing alters the reality being observed. The Carr principle (which I came up with this morning while eating breakfast) states: the act of searching alters the reality being searched.

The first web search engines based their results either on recommendations submitted by surfers or on the text of web pages. But as soon as a lot of folks started searching, these signals became corrupted. Site owners, seeing the commercial value of high search rankings, started to game the system. They flooded the recommendation systems with self-serving recommendations and they loaded their pages up with junk text written to push the pages up higher in the text-based rankings. (Remember those long stretches of repeated phrases tacked on to the ends of pages?)

Google came up with the more sophisticated idea of using links as signals of page quality. It worked great for a while. But then it spawned an entire “search engine optimization” industry bent on gaming the link system. The corruption of links forced Google to start tracking all sorts of other signals in hopes of staying ahead of the SEO goons and their corporate patrons.

Now, Facebook has introduced what it calls Graph Search. One of the main signals that Facebook is using to rank results is, not surprisingly, the “Likes” that it tracks via the Like buttons and other links it has spread across the web like so many dandelion seeds. Search companies in the past usually tried to choose uncorrupted signals as the criteria for their rankings. They wanted to give good, objective results in order to attract users. The corruption of the signals came later, after it became clear that the search results had commercial value. Facebook is taking a different tack. It’s starting with a signal—Likes—that is already corrupted, that in fact has always been corrupted. People routinely Like a thing not because they actually like it, not because they have (to use a favorite Facebook word) any real affiliation with it, but because they’ve been, in one way or another, bribed to Like it.

Like us on Facebook to download our new single! Like us on Facebook to get 10% off your next purchase! Like us on Facebook to get a chapter of our new e-book for free! Like us on Facebook to enter our sweepstakes! Like us on Facebook so our dad will give us a puppy!

Facebook never wanted Likes to be objective indicators of real affection, or even a vague feeling of fondness. The Like button was designed as a marketing tool, as Steve Cheney explains:

Early on FB made the case to brands that they must have fans… together with the ad agencies they convinced the Cokes of the world to spend money to be competitive (hey Pepsi is here too). Then, FB promised, something miraculous would happen.  Your friends would see in their news feed you liked Coke! So… FB convinced big advertisers to spend huge sums on CPA-like ad units whose sole purpose was to acquire fans. Ad agencies dedicated creative, planning and strategy resources to get the Cokes and American Expresses of the world to pay to have users click—almost 100% of the time because the user was promised some sweepstake or contest.

Even the normally decorous New Yorker got in on the act:


If you can’t read that, it says: “You must like The New Yorker to read the full text.” And some 17,000 Facebookers dutifully clicked the Like button. Jonathan Franzen must have been thrilled to see his essay used as a worm to bait a rusty Facebook hook.

It might seem kind of strange for a company to build a search engine — a pretty costly undertaking — using criteria that it knows to be debased, to be anything but objective. But to Facebook, it’s business-as-usual. Here’s the difference between Google and Facebook: Larry Page recognized that commercial corruption was a threat to his ideal. For Mark Zuckerberg, commercial corruption is the ideal.

Now, to be fair, corruption is not the same as absolute corruption. A corrupted search engine can still be immensely useful, as Google shows us every day. And a lot of Facebook Likes are actually likes. To be even fairer, Likes are not the only signal that is determining Graph Search’s results, and some of the other signals are probably, at the moment, purer indicators of affiliation and relevance. But you can bet a million Likes that the SEOers are already hard at work deciphering all those signals and their weightings in hopes of gaming the system. And they will succeed. If you see “social” as an antidote or counterweight to “commercial” on the web, the arrival of Graph Search should make your hair stand on end.

And, yes, the kids got their puppy.

11 thoughts on “Facebook’s polluted graph

  1. deniz oktar

    Thats right up to a point,

    In practice, most people wouldn’t like to search ‘friends who liked NY Times’ or ‘Friends who are fans of Pepsi’.

    99% of searches would be things like ‘friends who are single’ or ‘friends that live in new york’ or best of two worlds, imagine you are going to NYC for a trip, ‘friends that are single and lives in new york’

    Other than that, I don’t see facebook search going anywhere beyond in 3-5 years. What they are doing is in fact called semantic search, having all these likes and opengraph actions connecting. At one point of time, having the right connection between things, you might ask something like ‘find friends who have a motorcycle’ and anyone who had a facebook action such as ‘I ride a Piaggio’ would come up since facebook could manage to come up with an ontology that says ‘piaggio is a motorcycle brand’. This has some-how huge pottential however, we as end users do not oftenly need that much structured information. Mıst people won’t use it and it doesn’t bring any real solutions to real world problems we have right now.

  2. Evan Morris

    “Jonathan Franzen must have been thrilled to see his essay used as a worm to bait a rusty Facebook hook.” Ha! All human relations – and culture – subsumed into the cash nexus. I literally grew up reading the New Yorker (and pestering my mother to explain the cartoons). Now it’s more and more like reading the Style section of the Times, and that ain’t good.

    I have an unattended FB page automatically fed with teasers from my web site (word-detective.com). In the past year 2,134 people have “liked” it. I never go there because FB gives me the creeps. There’s something deeply wrong about it.

  3. Salman

    People often assume that Google and Facebook have come close to collecting all the world’s knowledge. This is especially the case, the chorus goes, since people are compulsive and uninhibited on the web, and therefore more liable to reveal “the truth” about themselves.

    I protest against this view of human knowledge–against its stupidity and its philistinism. The great resource of world literature teaches us nothing if not that we are most intimately attuned to ourselves when we take the time and care to meditate upon the subtleties and nuances of our subjective experience.

    Facebook’s Social Graph is therefore severely limited, and necessarily so. It is precisely that realm of our humanity that is most mysterious to us that our search engines fail to penetrate.

    Marcel Proust, that great man of solitude and introspection, suggests that “a very slight degree of self-acquaintance” teaches us that “a book is the product of a different self from the self we manifest in our habits, in our social life, in our vices. If we would try to understand that particular self, it is by searching our own bosoms, and trying to reconstruct it there, that we may arrive at it.”

    “In fact,” Proust writes, “it is the secretions of one’s innermost self, written in solitude and for oneself alone that one gives to the public. What one bestows on private life – in conversation…or in those drawing-room essays that are scarcely more than conversation in print – is the product of a quite superficial self, not of the innermost self which one can only recover by putting aside the world and the self that frequents the world.”

    “The beautiful things we shall write if we have talent,” Proust says, “are inside us, indistinct, like the memory of a melody which delights us though we are unable to recapture its outline. Those who are obsessed by this blurred memory of truths they have never known are the men who are gifted … Talent is like a sort of memory which will enable them finally to bring this indistinct music closer to them, to hear it clearly, to note it down …”

    If we allow Facebook to dominate our lives, and to become the primary medium of discourse, we impoverish out culture immeasurably. But my suspicion is that the more we trumpet the social network as a revolutionary force for transparency and unification, the more its shallowness and poverty will make itself apparent.

    The combination of naked greed and cheerful utopianism, which characterizes Facebook and Google, and which has characterized so many dangerous and creepy regimes in the past, is not indefinitely sustainable.

    Some things can’t be easily ignored, even by those who have announced their intention to make the world a better place, and to do so at the cost of our art and our individuality.

  4. David Schmaltz

    Lovely. Seems likely (pun intended) that one day soon, the ‘industry’ supported by the presumed relevance of likes and search engine popularity will prove to be another bubble incapable of generating sufficient revenue for those paying for ads to continue justifying the outlay. It was a fine, promising, and probably vacuous notion in the first place.

  5. Kobra Koskinen

    Excellent post, I have wrestled with the same questions before: how to define quality of a page in a way that wouldn’t be corrupted. Closest that I have gotten is taking a look on how people interact with the content; it’s not perfect, but it’s a lot more difficult to hack than like buttons or links. We’re starting with online articles, but this type of perceived quality evaluation would be aplicable to other content as well. http://venturevillage.eu/scoopinion-launches-leading-online-migration-toward-good-reads

  6. Mathew Ingram

    I take your larger point, Nick — but that’s not the Heisenberg uncertainty principle you’re referring to, it’s the “observer effect,” which is a somewhat different animal.

  7. Mathew Ingram

    I expect to see a full correction, complete with an update describing the difference between the two as it pertains to your conclusions :-)

  8. MikeD

    I’m surprised that only Facebook-hosted content will be searchable. There will be no links to other sites essentially. With the Like button scattered across the Web, merely visiting a page provides a signal to Facebook, regardless of whether you click the button or not. This gives FB information about which sites are really being visited and what kind of audience is interested – you’d think that would be useful somewhere.

Comments are closed.