Scathing critique of "social" sites: "The Social Graph is Neither"

Maciej Ceglowski's "The Social Graph is Neither" is a scathing, spot-on critique of the deceptive and seductive simplicity of "social graphs" which purport to represent human interaction and relations through mathematical modelling. As with many "semantic web" projects, social networks can only achieve any kind of usable scale and coherence by simplifying the relationships they model to the point of triviality.

One big sticking point is privacy. Do I really want to find out that my pastor and I share the same dominatrix? If not, then who is going to be in charge of maintaining all the access control lists for every node and edge so that some information is not shared? You can either have a decentralized, communally owned social graph (like Fitzpatrick envisioned) or good privacy controls, but not the two together.

There's another fundamental problem in that a graph is a static thing, with no concept of time. Real life relationships are a shared history, but in the social graph they're just a single connection. My friend from ten years ago has the same relationship to me as the friend I dined with yesterday. You're left with forcing people (or their software) to maintain lists like 'Recent Contacts' because there is no place in the model to fit this information.

"No problem," says Poindexter. "We'll add a time series of state transitions and exponentially decaying edge weights, model group dynamics as directional flows, and pass a context object in with each query..." and around we go. p> This obsession with modeling has led us into a social version of the Uncanny Valley, that weird phenomenon from computer graphics where the more faithfully you try to represent something human, the creepier it becomes. As the model becomes more expressive, we really start to notice the places where it fails. p> Personally, I think finding an adequate data model for the totality of interpersonal connections is an AI-hard problem. But even if you disagree, it's clear that a plain old graph is not going to cut it.

Pinboard Blog (via O'Reilly Radar)

(Image: Map of top 50 UK PR twitter people and their followers, a Creative Commons Attribution Share-Alike (2.0) image from porternovelli's photostream)


  1. Nice use of “Poindexter”.  A sadly neglected “Felix The Cat” character: so much so that some earlier versions of the OED, before the intervention of the “Friends Of Felix”, incorrectly attributed this meme to a much later media occurrence.  

    1. I always go to my pastor when I need help finding a good dominatrix… and if he cannot help, to my kids kindergarten teacher.

  2. Honestly, I find this sort of argument against mathematical modeling to be dismissive and lazy. No, we can’t provide all the details of most complex systems by models alone. But does it matter? Even if it’s possible to capture all the details of a system with a complex model, that’s not the point of most models anyway. Many modeling approaches, especially those with a solid analytical basis, are about distilling complex interactions into something more tractable where we can learn about fundamental behaviors or drivers of the system that might not be apparent otherwise. Focusing on particular aspects of a system isn’t “creepy,” it’s potentially illuminating, and it’s the basis for most scientific endeavor.

    I’m dismayed to hear my colleagues (microbiologists) often dismissing mathematical models as being too “simple,” or complaining that they strip the “beauty” away from the real world. Models don’t aim to be complete representations (why should they be when we can look at the actual, complete system when that happens to be our focus?), and they can provide perspective on the true beauty in the world by helping us to wrap our minds around it.

    1. Or as the statistician George Box wrote: “Remember that all models are wrong; the practical question is how wrong do they have to be to not be useful.” (sometimes simplified to “all models are wrong but some are useful”)

    2. I am in almost complete agreement with the above comment. I would like to add this: If you feel that  these models trivialize social interactions, please provide an alternative. Something objective. Something that makes predictions. Something that at least has the virtue of being wrong. 

  3. I’ve never bought the OMG! Social networks can expose your vices (use of dominatrix, drinking, whatever) hysteria. People should decide what sort of entertainments they like and if these go against whatever prudish moral code they claim to uphold, that is a really good sign that they they ought to decide which is more important to them — the entertainments or the code, and adjust one to meet the requirements of the other.

  4. “…simplifying the relationships they model to the point of triviality.”Isn’t this a fairly accurate definition of what a model is?

    “…the Uncanny Valley, that weird phenomenon from computer graphics…”
    I believe this term/concept predates CGI by quite a bit. Wikipedia has this to say: “The term was coined by the robotics professor Masahiro Mori as Bukimi no Tani Genshō (不気味の谷現象) in 1970”

    Would it be wrong to say this author blablabla to the point of triviality?

  5. Can your computer do EVERYTHING?
    No. Do you complain about that?
    Can social networks do EVERYTHING?
    Of course not, you fool.
    But can they be used, by trained social scientists, to perform rigorous analyses, taking directly into account every one and more of the obvious shortcomings you pretend nobody else has thought of, of behavioral phenomena that, in conjunction with the computational power available today, shed significant light on processes of information availability, job procurement, educational attainment, physical health, political beliefs, and I can’t read everything from the last 60 years to make this list exhaustive?
    Uh, yes. Yes, they can.

    I’m sorry—social graphs come from sociology. The substance of this ten-page strawman can be compressed to—People who aren’t trained as sociologists aren’t as good at sociology, and modeling things is too tempting to avoid, despite being hard. Which are valid points. But also, dare I say it, “trivial”?

    (I don’t restrict the response to online networks because this is an indictment of network methodology in general, just fueled by examples of bad science from online. But the notions that networks have to be studied online, or that there can’t be digitized IRL networks, are both absurd. In particular, because they’ve both already happened—one about 6 years ago, the other about 60.)

  6. Having mostly read commentary today on the death of Flash which as you can imagine was largely religious-war drivel, its truly refreshing to see that there is in fact a place where thoughtful, interesting comments can be found…

  7. I’m getting the sense thus far that the complete text is not getting read… He has been actively involved in the guts of social media and acknowledges that there is utility to the likes of sociologists and the like in representing social relationships as graphs. He’s not dim. What he’s primarily noting is that what the likes of Facebook do is the reverse- trying to cram real-life social relations back into the little toy model creates something wholly unlike, and sometimes downright hostile to, real world relationships, but really friendly to people trying to sell you shit, and that’s an observation I think anyone whose has gotten over the new platform honeymoon period has concluded and they’ve seen all the envy-inducing vacation photos and got their first friend request from their weird office mate.  The only social relationship we feel the need to publicly establish past the age of three are marriages, and yet we have this goofy framework for taking every human interaction, stripping them of distinction (is there a dropdown for selecting “long standing unrequited crush you having a good working relationship with but would rather not have your cousin know you still talk to?”,) and then making them public, which is itself a social action, and the resulting weirdness of goldplating and mummifying often tenuous relationships, and serve up the resulting flood of personal intelligence to advertisers, who have…unfortunate tendencies.  It’s just odd. It’s what happens when you sit down with the kids in the dorm who were trying to work out dating prospects with a spreadsheet and proceed to be baffled why they don’t get dates.

  8. The only rational way to look at social networking platforms is in the same as any other software: tools which serve a purpose.

    Some people use LinkedIn as an address book, some to find jobs, some to do research, find business opportunities, etc. Facebook is a source of engagement and entertainment for many people. Very many people. Obviously. So it’s a pretty useful tool which serves its purpose well.

    I seriously doubt that Facebook (for instance) make any pretenses at accurately reflecting something as complex as human relationships. What they *do* have is a useful abstraction thereof, which fits a fairly narrow slice of the full spectrum of use cases. Other social networking sites focus on a different part of the spectrum. There are dating sites, Twitter, Enterprise 2.0 and Social Business, Social CRM tools, the list goes on.

    Social networking platforms implement abstractions which are only true insofar as they are useful. Period. The same can be said for any model.

    The argument in the article is built taking the word “social” out of its context. The “social” in “social graph” is understood to mean “a digital representation of a person and their relationships”. Nobody pretends that the node *is* the person. Just a representation. Why should anyone pretend that the edges connecting any such representations are anything more than representations themselves?

    So, given that it’s software; given that it’s an abstraction, and given the context; social graphs are certainly both.

Comments are closed.