June 8, 2010

The most important part of the “social graph” is neither social nor a graph

“Social graph” is a highly misleading term, and so is “social network analysis.” By this I mean:

There’s something akin to “social graphs” and “social network analysis” that is more or less worthy of all the current hype – but graphs and network analysis are only a minor part of the whole story.

In particular, the most important parts of the Facebook “social graph” are neither social nor a graph. Rather, what’s really important is an aggregate Profile of Revealed Preferences, of which person-to-person connections or other things best modeled by a graph play only a small part.

Let me hasten to note that – even when viewed narrowly — the ideas of “social graph”and “social network analysis” do have significance. Nontrivial use cases to date for big data social network analysis include:

Epidemiologists aspire to add to that list, based on their success to date using much more micro forms of social network analysis. But after that, I’m running out of examples. Sure, graph analytics is good for a bunch of other things (e.g., biology at the genetic or molecular level), but those have little or nothing to do with “social graphs” or social network analysis as they are commonly understood.

Note: Of course, it is also the case that everything can be modeled by entity-attribute-value triples, and those can always be modeled by graphs. But so what?

Let’s consider what, in a marketer’s ideal world, would go into your Profile of Revealed Preferences. Raw data might include:

My core privacy thoughts about that data include:

When one frame things this way, two rather difficult technological questions naturally arise.

  1. Suppose, implausibly, that a single entity were allowed to control and use (for marketing) all of your Profile of Revealed Preferences information. How would they store and analyze it?
  2. How does the answer to #1 change because control over the information will, in fact, be fragmented?

It’s tough enough to answer these questions for data about one person. Trying to include all but the simplest information about other people is and will for years remain quite infeasible. So, for the most part, this is not “social” information.

It’s also not naturally a “graph.” Similarly, it is not a good candidate for network analysis. To see why, let me outline why I used the name “Profile of Revealed Preferences”:

Not much graph-like there.

This post has gotten pretty long, so I’ll stop here without spelling anything else out. But questions I still hope to address down the road include:


13 Responses to “The most important part of the “social graph” is neither social nor a graph”

  1. J. Andrew Rogers on June 8th, 2010 3:24 am

    Great topic with several interesting tangents that could be written on at length. A few semi-random thoughts:

    – The real value of graphs is that they can be a universal representation and access method in theory. You can seamlessly mix-and-match your data with anyone else’s. In practice, typical graph implementations scale so badly that rigid, non-universal representations and access methods are more attractive.

    – Almost no one does true graph analysis due to the difficulty of scaling operations like transitive closures. I could easily add another half-dozen industries to your list that badly want true graph analytics but can’t get the scale to make it economical. Graphs tend to devolve to a key-value store as scale requirements grow.

    – The idea of “graphs” in mathematics has a broader scope and is more exotic (and powerful) than the simple point-link-point model ubiquitously implemented. It is easy to see why this simple model is used since the more exotic models are not visualizable; you have to retreat to “data structure design by obscure mathematics” that is difficult to reason about.

    – An aspect of graphs that is not immediately obvious is that they have a direct relationship to algorithmic information theory that allows very powerful types of computational induction that can pull potent patterns out of surprisingly diffuse bits. However, this type of use case is provably intractable using trivial graph constructs (but not for exotic ones). Most people working with social graphs intuit the possibility of something like this even if unfamiliar with the mathematics.

    – If someone really figures out the “induction over exotic graphs” angle, the privacy arguments almost become moot because sufficiently competent induction can reconstruct most of it from latent bits of we unavoidably leak everywhere in our lives. Even at the level of mathematics published now, the ability to reconstruct entities from diffuse environmental bits has been getting very good, very quickly. Most people have not thought about the implications of this, they are assuming that personal information is aggregated via conventional record sharing channels rather than exotic information theoretic reconstruction. Still a mostly correct intuition but not for long.

    Social graphs are boring in large part because doing something not boring and scalable is very non-trivial. A lot of the social graph companies are hoping to be sitting on the killer data set as it gets solved but they aren’t considering the implication of my last point and the fact that all services are inherently leaky.

    This comment became much too long…

  2. J. Andrew Rogers on June 8th, 2010 3:46 am

    Let me add that while I’m using graphs generically here, I’ve seen incredibly detailed social graph models that go far beyond the friend-of-a-friend case. When trying to develop deep contextual models, the social graphs become complicated. The interactions between two individuals is naturally very dynamic and contextual, so for the kinds of behavioral prediction organizations are interested in this is important.

    A really good social behavior model is so graph-like that it is intractable for anything useful even at small scales. Yeah, we get to see the “FoaF” model as consumers, but the internal models are often accumulating more detail than that about the influence of network dynamics on behavior.

  3. Alan on June 8th, 2010 4:07 am

    ‘But after that, I’m running out of examples.’ Two examples I’ve come across.

    Preferential customer service – similar to the ‘Telephone companies use calling data’. Based on estimated net worth of the social network, a fin services firm goes to extra-ordinary lengths to deliver customer service. You never know who will marry a Kennedy these days – and they *all* talk.

    Targeted promotions. Think travel for ‘Social networks figure out which other members you’re likely to know’. Based on past history and geography of the social network there’s a limited time offer for discounted travel rates. This isn’t new, it’s just one more source of data for targeting.

  4. Dan Weinreb on June 12th, 2010 1:04 pm

    And think of how political operatives can take advantage of this information, sending different candidate pitches (or different misleading phone calls!) based on all of these attributes. (Facebook even asks for your political affiation.)

  5. Evan on June 16th, 2010 12:50 am

    Seriously good points. The visualisation of these networks is great fun (and often fascinating), but apart from direct case management and investigation in fraud or intelligence, I must admit I’ve seen few situations that wouldn’t have benefited as much (if not more) by simply treating personal network characteristics and behavioural patterns as additional inputs into a model of some form. Plus, modelling entire populations is often less interesting than modelling subpopulations of tightly connected individuals – the intuitive response seems often to be to try and understand *everything*, but the reality is that I typically care less of your “Friend of a Friend of a Friend” connectivity than I do about your personal behavioural characteristics (which I may either be trying to influence or be viewing with suspicion, depending on why I’m looking at you in the first place). You may be seven degrees removed from Kevin Bacon, but given the amount of time it’ll take for a message to propagate from you to Kevin, that connectivity is pretty much useless.

    We’re visual creatures and the graphs are easy to latch onto as something tangible and attractive, but in practice I’ve found they’re the least interesting part. Essential, but not so interesting compared to what’s possible.

  6. Objectivity Infinite Graph | DBMS2 -- DataBase Management System Services on June 19th, 2010 8:27 am

    […] Darren is in general agreement with my observation that the “social graph” shouldn’t primarily be regarded as a graph. […]

  7. Social Technology on June 22nd, 2010 6:31 am

    Difficult Chapter Delayed Me, Difficult Data, Too…

    I found your entry interesting thus I’ve added a Trackback to it on my weblog :)…

  8. Data that is derived, augmented, enhanced, adjusted, or cooked | DBMS 2 : DataBase Management System Services on November 29th, 2010 4:57 am

    […] kinds of predictive analytic models, real-time scoring could be prohibitively slow. Ditto for social graph analysis, and the same goes for the other examples as […]

  9. The technology of privacy threats | DBMS 2 : DataBase Management System Services on January 14th, 2011 9:14 am

    […] graph” it infers from informal communication among friends. To date that assumption has been questionable, but we’re still in the very early days. Meanwhile, cruder methods of analyzing social […]

  10. Big Data is Watching You! | DBMS 2 : DataBase Management System Services on January 22nd, 2011 3:45 pm

    […] that they have enough of a social graph to do so. (This application is a case where the term “social graph” is not a […]

  11. What to do about “unstructured data” | DBMS 2 : DataBase Management System Services on May 16th, 2011 7:03 am

    […] In June, 2010 I elucidated the variety of data that could go into an individual’s marketing-oriented profile. […]

  12. Think clusters, not networks, in mining the social graph. on August 31st, 2011 12:22 pm

    […] into this kind of data, and use the best available analytics to get at what some pundits call the Profile of Revealed Preferences, among other terms.  That involves not just their connections, but their preferences and behaviors, […]

  13. Think clusters, not networks, in mining the social graph. | Volvant on August 31st, 2011 2:16 pm

    […] into this kind of data, and use the best available analytics to get at what some pundits call the Profile of Revealed Preferences, among other terms.  That involves not just their connections, but their preferences and behaviors, […]

Leave a Reply

Feed: DBMS (database management system), DW (data warehousing), BI (business intelligence), and analytics technology Subscribe to the Monash Research feed via RSS or email:


Search our blogs and white papers

Monash Research blogs

User consulting

Building a short list? Refining your strategic plan? We can help.

Vendor advisory

We tell vendors what's happening -- and, more important, what they should do about it.

Monash Research highlights

Learn about white papers, webcasts, and blog highlights, by RSS or email.