e-Literate

Present is Prologue

Tag: scalefree-networks

  • Homeland Security, Networks, and Accountability

    Social networks guru Valdis Krebs has written an analysis which purports to show that creating a single intelligence czar is less efficient in terms of getting intelligence to the President than leaving the stove pipes but adding connections among them.

    Maybe.

    But Krebs leaves out at least two critical externalities. First, his proposed solution requires that no less than five intelligence directors report directly to the President. Meanwhile, State, Treasury, Defense, and so on each only have one. Ain’t gonna happen. No matter how important intelligence is, the President has other priorities to deal with as it is. So it would seem that the compromise Krebs raises and dismisses of connecting the stovepipes under the intelligence czar would be the best solution.

    Or would it? The second problem is that Krebs’ map fails to recognize that there is not just one network but (at least) two that we need to be concerned with. He has mapped knowledge flows. But there is a different network of budget, command, and control. This matters in several ways. First, as the 9/11 Commission pointed out, if you can’t hire, fire, and set budget priorities, then the information that flows to you may not be the information that you need. Furthermore, in a representative democracy, there needs to be accountability. You need somebody who can say, “The buck stops here.” Self-organizing systems may bring efficiency (when they work), but it’s not at all clear that they help accountability. If you add to this the fact that the largest player in the intelligence game is the Pentagon which, for understandable reasons, has a very strong command and control hierarchy, it becomes clear that networking the silos is problematic. The lines of communication can easily run afoul of the lines of command-and-control.

    In an ideal world, you want to either completely harmonize the two networks or completely separate them. Within a command-and-control structure, you can set up parallel siloes and tight networks of communication–as long as there is somebody at the top with the authority to force alignment of goals and be ultimately accountable for organizational failures. When you have to cross organizational lines, it’s best to do so in a strictly advisory capacity with clean lines of command-and-control separation. From the little information we have on Colin Powell’s proposal to create an intelligence advisory committee similar to the Joint Chiefs of Staff, we don’t know enough yet to say whether this fits the bill. It might; the Joint Chiefs of Staff are removed from the chain of command, so there’s an indication that Powell is trying to disentangle the knowledge network from the accountability hierarchy. What we don’t know yet is how the vast majority of the intelligence agents will be organized. Will they remain in their current agencies or will they be centralized under the authority of the intelligence czar? I fear that the former would not do the trick. However, the latter might enable the stovepipes to remain intact under a new accountability hierarchy that is not in tension with the knowledge network.

    We’ll see.

    (Found via ScaleFree.net.)

  • It's a Small Campus After All

    Gilad Ravid and Sheizaf Rafaeli’s new piece in FirstMonday, “Asynchronous Discussion Groups as Small World and Scale Free Networks“, analyzes a voluntary learning community that develops on a university’s LMS. These are all students who are (apparently) registered for on-campus web-enhanced courses with strictly voluntary web-enhanced components. Interestingly, the study analyzed networking for the entire group of online participants across course boundaries. So the results are for the network characteristics of the entire online student body, as opposed to the network characteristics of student participation in individual courses. Ravid and Rafaeli find that the community formed by the students appears to be both a Small World and a Scale Free network.

    What does this mean? (more…)

  • overstated: Weblogs and authority

    In the Overstated weblog (great name, by the way), Cameron Marlow suggests that blogrolls are proxies for popularity while links directly from a blog post to a permalink of another blog are proxies for influence. For example, slashdot is popular in blogrolls but Joi Ito is popular to link to in posts. Marlow does some empirical analysis showing that, while some of the very top sites show up in both permalinks and blogrolls, the two groups (i.e., the group of heavily permalinked blogs and the group of heavily blogrolled blogs) diverge dramatically in the lower ranks.

    Why? Marlow doesn’t address that question in his post, but my guess is that it has to do with competitive differentiation. Bloggers who are serious about building their reputations need to find content that nobody else in their circle is talking about. If they only got their material from the hubs that everybody reads, then they would have little to make them stand out from the crowd (except, perhaps, quality of analysis and quality of writing). So competition for readership may be the reason why bloggers tend to link to sites that are not on everyones blogroll yet. (By the way, as Marlow correctly points out, this effect tends to mitigate the power law problem.

  • Weblog Audience-Building and the Strength of Weak Ties

    One of the challenges you face when you start a new weblog is attracting an audience. Who is going to gather the pearls of wisdom that you offer to the world? It’s not that hard these days to find somebody you know who already has a weblog and would be willing to link to you; at the very least, by now your mom probably has a weblog and would be willing to do you a favor. But what then? How do you grow your audience beyond your circle of acquaintances? How do you get people who don’t know you to link to you?

    Sociologist Mark Granovetter provides a clue in his oft-quoted classic paper “The Strength of Weak Ties.” (more…)

  • Informational Cascades, Network Theory, and Behavioral Economics

    Stephen Downes’ mention of my article on informational cascades (thanks for the plug, Stephen) led me to his post in the trdev discussion group. He writes:

    In network theory, ‘groupthink’ is an instance of what is known as a cascade phenomenon. A cascade occurs (all other things being equal) when the propogation of a property (an idea, a software acquisition, a disease) exceeds 1 – that is, each instance of the phenomenon replicates on average at least one time. It is important to note that the political organization of the group is irrelevant – that’s why mixing ideology with the study of groups can be dangerous.

    Classic examples of this principle include stock market bubbles, ranging from the tulip mania that swept through Holland in the 1630s to the Tech Bubble we just experienced to the widespread adoption of email to the influenza epidemic of 1919.

    While Stephen is correct to point out the interest that network researchers have in cascades, and while some network researchers are interested in informational cascades in particular (for example, Duncan Watts has a small section on them in Six Degrees), it may be worthwhile to point out that informational cascades come from a different intellectual lineage.

    For many decades now, economics has been dominated by the “rational choice” model of human behavior. This basically says that, when making a decision, humans look at how likely it is that that a given option will produce the desired outcome and the expected utility of that outcome. People (according to the model) choose the option that is most likely to produce the best outcome.

    Except there are lots of cases where people clearly don’t perform these calculations as the rules of logic and statistics would require. They are doing some kind of calculations, but not exactly the ones that we would expect from the rational choice model. So a school of thought has developed within economics and psychonology that suggests that humans use mental shortcuts to make approximations of those utility calculations. Often those shortcuts will lead to a “good enough” answer with much less time and mental effort, but occasionally they lead people astray. This model, called the “bounded rationality” model, is the heritage from which informational cascade theory comes. Listening to the advice of trusted friends and colleagues is one of those short cuts, and it is the essential mechanism that leads to informational cascades. (A whole list of these shortcuts is being researched in the field of psychology called “heuristics and biases.” There’s also a rich literature on it within behavioral economics.)

    At any rate, what’s interesting is that informational cascades happen to map so well to network theory despite the fact that they come from a totally different research thread. In my view, one of the more fertile and interesting ways to start figuring out what kinds of wild and whacky things can happen in online knowledge sharing is to overlay heuristics and biases theory on top of network theory.

  • Is Johnson's "Clustering Emergence" Really Small-world Network Formation?

    I was thinking some more last night about Stephen Johnson’s new position that there are separate types of clustering and adaptive emergence as I was reading Albert-Laszlo Barabasi’s book Linked (which I am enjoying immensely, by the way; more on that in a later post). I suddenly had a flash of intuition that what Johnson now refers to as “clustering” emergence in the Dean campaign and other political grassroots events may just be the formation of a small-world network.

    Here’s the evidence:

    • The Dean campaign achieved incredibly rich, fast, and robust communication with its grass roots, just as you would expect in a small-world network.
    • The Blog For America site acted as a network hub, which is just the kind of catalyst you need to turn a normal grassroots network into a small-world network.
    • The sense that this powerful, organic “system” suddenly “emerged” is consistent with the kind of sudden and dramatic phase transition that is typical at the moment of small-world network formation.

    Intriguing, huh?

    It might be interesting to graph the Dean campaign’s network of support web sites, blogs, etc., and compare it to the other campaigns. Did Dean’s support web conform to a power law? Did Kerry’s? Bush’s? One aspect that makes this harder is that, with MeetUp factored into the picture, the network extends outside the Internet in some significant ways and therefore might be a lot harder to map.

  • Link Prediction in Social Networks

    I’m afraid this is going to be something of an echo-blog, since I don’t have the paid subscription to the ACM library necessary to allow me to see the original article, but the About Kim weblog has an intriguing and encouraging quote from an academic article on link predictability within social networks:

    By running our predictors on some other datasets, we have discovered that performance swells dramatically as the topical focus of the dataset widens. In a narrow field, almost anyone can collaborate with anyone else, and new collaborations are largely random. It would be interesting to make precise a sense in which such new collaborations are simply not predictable from the training data.

    From this blurb it looks like there exist conditions within certain social networks in which past collaboration or references are not highly predictive of future collaboration or references. In other words, people make new connections with new people a lot.

    I’m going to have to get my hands on this article; it’s too tantalizing.