Diffusion of Informational Contents

by Jacob Lee

Research into diffusion processes permeates disciplines as diverse as computer science, anthropology, sociology, economics, epidemiology, chemistry, and physics[1]. Much recent work, in the last fifty years or so, has been explicitly network oriented and has sought to better understand how network topology and transmission mechanisms determine properties such as the rate of diffusion and the various thresholds at which diffusion processes become self-sustaining.

Despite the many apparent similarities between different diffusion processes, it is important to be attentive to the particulars of each kind of diffusion process. Commercial products diffuse among consumers in ways different than do news articles or news topics in the blogosphere (see Dynamics of the News Cycle). Behaviors like smoking spread across networks of friends in both similar and contrasting ways as those of sexually transmitted diseases. And routing information in a sensor network propagates differently than routing information in a mobile phone network.

Diffusion of Semantic Content

The diffusion of information or more generally semantic content has been a cross-disciplinary concern, and has been treated in a variety of ways, depending on the domain of application. It is generally recognized that such content exhibits properties that distinguishes it from the diffusion of other phenomena. For example, it is recognized that the sharing of semantic content, unlike commodities, does not necessarily incur a consequent loss of that content for the sharer, and that information is often shared preferentially with those for whom it may be of interest or desired[2].

Nonetheless, in more general settings the implications of the properties of semantic content for its diffusion has to my knowledge not yet been formally investigated. Content is typically treated as a non-relational item whose diffusion-mechanism is essentially content-neutral, except perhaps in its differential transmissibility or mutability. Furthermore, it frequently restricts its models to the diffusion of isolated pieces of content in an otherwise content-less context. Consequently, it confounds the diffusion of content vehicles with the diffusion of the semantic content itself, treating content vehicles as having an intrinsic meaning or significance.

It is true that the transmission of content vehicles is easier to understand, and that this simple approach probably does a fair job of approximating the diffusion of semantic content at a unit of content or level of abstraction at which the applicability of a more rigorous approach may not be either readily apparent or especially necessary. Yet it has the unfortunate effect of potentially blinding us to the way in which the relation between contents and cognition (or computation) can generate a second, more leaky, means of content diffusion, or can inhibit or transform content. For example, it is entirely possible for  multiple agents in a network to independently infer the same piece of information without that  piece of information ever having been explicitly communicated to them. It is also, I might add, entirely possible for two agents to receive the same communications and infer entirely different things, or fail to interpret the message correctly, as anyone who has had to try to collaborate with others by email can readily attest.

[1] See my recent blog post Processes of Diffusion in Networks

[2] Unlike disease, which neither the giver or receiver desires!

Advertisements

17 Comments to “Diffusion of Informational Contents”

  1. Jacob, there is an article in the Volume 32, Issue 4, October 2010 issue of Social Networks that seems relevant to this topic. “Social creativity as a function of agent cognition and network properties: A computer model” by Siddhartha Bhattacharya and Stellan Ohlsson, pp. 263-278. The abstract reads as follows,

    “Inventions — concepts, devices, procedures — are often created by networks of interacting agents in which the agents can be individuals (as in a scientific discipline) or they can themselves be collectives (as in firms interacting in a market). Different collectives create and invent at different rates. It is plausible that the rate of invention is jointly determined by properties of the agents (e.g., their cognitive capacity) and by properties of the network interactions (e.g., the density of the communication links), but little is known about such two-level interactions. We present an agent-based model of social creativity in which the individual agent captures key features of the human cognitive architecture derived from cognitive psychology, and the interactions are modeled by agents exchanging partial results of their symbolic processing of task information. We investigated the effect of agent and network properties on rates of invention and diffusion in the network via systematic parameter variations. Simulation runs show, among other results, that (a) the simulation exhibits network effects, i.e., the model captures the beneficial effect of collaboration; (b) the density of connections produces diminishing returns in term of the benefits on the invention rate; and (c) limits on the cognitive capacity of the individual agents have the counterintuitive consequence of focusing their efforts. Limitations and relations to other computer simulation models of creative collectives are discussed.”

  2. Fascinating. Could we study the diffusion of new concepts by language users at particular moments (e.g., “social network” or “death panels”) within this kind of framework?

  3. That looks like an interesting paper, John. There is also some recent work done on so-called epistemic communities that is a combination of network theory with formal concept analysis.

    Dave: there has been work doing something like that. I write about it here: http://jacoblee.net/occamseraser/2010/08/06/dynamics-of-the-news-cycle/ You can take a look at Meme Tracker http://memetracker.org/

  4. Dave, Jacob beat me to it and has far more technical expertise in this area than I have. I would add that the early research on this problem, e.g., a study of the diffusion of an educational innovation in a network of high school principals, framed the problem in terms of influence and started with the idea that better connected people would be more influential, a thought that was picked up by marketers interested in opinion leaders. Jon Kleinberg, one of the creators of Meme Tracker, is the author of a famous article on techniques for identifying hubs and authorities on the Web. The key observation in the paper is that well connected hubs and authorities are not identical. Thus, for example, a term based search using a web crawler for a common term, e.g., “Harvard,” will inevitably wind up pointing to major portals, e.g., Google or Yahoo!, through which such searches are funneled. Authorities are, moreover, only a subset of those sites to which the big search engines point. The problem is how to identify that subset.

    The article in question is “Authoritative sources in a hyperlinked environment” by Jon M Kleinberg (1999), Journal of the ACM, Vol 46, Issue 5, pp. 604-632.

  5. One more thought. The graphic at the top of the Meme Tracker site covers only a three-month period in 2008, and, given the excitement generated by a presidential campaign and the 24-hour news cycle, it is not surprising that memes surge and decline very rapidly. It would be interesting to see what would happen with data covering longer time spans and a way to filter long waves from short ones.

  6. Thank you John, but I make no claim to any special expertise.

    Anyway, the Memetracker work is an interesting illustration of some of my main points. It is enlightening to see how families of related phrases propagate; presumably some kind of concept or idea is also propagating along with it. In many cases it is probably pretty clear that it is. But the conceptual content of these messages is hidden from view, because that content is, essentially, the result of the interpretation of these messages, and resides, at least in part, in the heads of people, and in how they are situated and used in other semantic content.

    So, on the one hand, tracking these “memes” tells us a lot, but it only tells us about what’s in the heads of people out there in as much as we have access to how that content is interpreted and understood by those people, and how it interacts with and relies on other content. It is not as if we can’t learning anything this way, but a great deal of caution is warranted. Do we not see material traditions leaping cultural boundaries so that while form is intact across both, it’s meaning or local significance is markedly changed?

  7. Thanks, Jacob and John. These references should be very helpful. I suspect that these kinds of studies of diffusion could go in historical directions as well (that’s where my research lies, since I’m an 18c lit scholar). The appearance of full-text databases of an 18c books and newspapers should make it possible to perform at least some searches for memes over longer periods of time. This is something that the field of conceptual history has tried to do historically, but the availability of masses of new primary source archives makes this kind of study possible. But I can imagine all sorts of interesting work done, for example, in book history as well.

  8. Jacob, I agree with you completely that studies like memetracker at best give you only indirect information. (like those studies of Victorian culture that count the number of times that “God” appears in 19c books) But I think even having an awareness of “flows” and “branchings” of various terms and phrases would have suggestive connections to other kinds of events. But book history has long had this problem, for example, of trying to evaluate whether someone’s ownership of a particular book meant some kind of familiarity with its content. Sometimes yes, sometimes no. I’ve been to rare book rooms where big volumes had to have their pages cut so that they could be read; the original owners, and subsequent library-users, never reached that part of the book.

  9. The first article I mentioned above is an attempt to use a two-level analysis to address the the relative impact of network factors and reception, processing, and transmission by the actors who are the nodes in the network. LIke any other model, it makes assumptions that simplify the analysis. While far from the last word on the subject, it is, it seems to me, a serious step into the territory described by Miller and Page in Complex Adaptive Systems: An Introduction to Computational Models of Social Life as lying between the two usual extremes, agents that are either too stupid (simply following prescribed heuristics) or too smart (calculating every move using game theory).

  10. It took me a while, but I’ve obtained a copy of that paper.

  11. The Kleinberg or the one from Social Networks?

  12. Bhattacharya and Ohlsson’s paper on social creativity. I’m afraid that I haven’t read it yet; I’ve a stack of other papers, unrelated to networks I’m afraid, necessary that I read at the moment. In time…

  13. I also have been unhappy with the availability of building materials, but I thought I was alone. Thanks, Jacob, for pointing out how widespread this problem truly is!

  14. Hm, didn’t mean to kill this thread with my silliness.

    “For example, it is recognized that the sharing of semantic content, unlike commodities, does not necessarily incur a consequent loss of that content for the sharer, and that information is often shared preferentially with those for whom it may be of interest or desired[2].”

    The costs tend to be in time and attention, on both sides.

    “Nonetheless, in more general settings the implications of the properties of semantic content for its diffusion has to my knowledge not yet been formally investigated. Content is typically treated as a non-relational item whose diffusion-mechanism is essentially content-neutral, except perhaps in its differential transmissibility or mutability.”

    I think you’re right to call attention to context variables and relationality. Otherwise, how could we explain students at a private university actively resisting the content of classes for which they pay thousands of dollars?

  15. I suppose that I ought to hedge the second statement you referred to and admit that in a broad sense the implications of content and context sensitive interpretive processes in the diffusion of information is examined in several fields, including communication studies, philosophy, logic, anthropology, sociology, etc… What I haven’t seen is these more nuanced understandings of information, meaning, and communication embedded in the framework of social network theory.

  16. What I haven’t seen is these more nuanced understandings of information, meaning, and communication embedded in the framework of social network theory.

    True, but things are beginning to move in these directions. Historically, social network analysis has focused on network structure and been motivated by the proposition that network structure might be an important independent variable in explanations of social behavior. More recently, work on the intersections of social and semantic networks has become a prominent theme, looking for correlations between clusters in semantic networks, where the nodes are terms or concepts, and social networks, where individuals or groups are the nodes. Another approach, prominent in epidemiological and other studies of diffusion, is to model variations in acceptance and transmission of whatever (disease, meme, etc.) is being transmitted. Much of the math in this area comes from the physics of phase changes, where, for example, the math used to model nuclear reactions has been applied to things like bar fights, the assumption being that the presence of of enough, properly distributed “dampers” will prevent heated debates from exploding into physical violence. The paper I mentioned earlier takes a further step, attempting to model the internal processing step between acceptance and retransmission, during which a mutation or change in meaning can occur. This work is, however, still in its infancy. As John H. Miller and Scott E. Page remark in Complex Adaptive Systems: An Introduction to Computational Models of Social Life, (Princeton: Princeton University Press, 2007), the agents in most social network models are either too stupid (following prescribed heuristics in a mechanical way) or too sophisticated (basing every decision on game-theoretic calculations). As far as I can make out, the people engaged in developing better models are still wrestling with the problem vividly described in my favorite article title “Artificial Intelligence Meets Natural Stupidity” (the author is AI researcher Drew McDermott. The article was written in the late 1970s; where it appeared, I am not sure).

  17. P.S. I should have searched the Web before saying that I didn’t know where “Artificial Intelligence Meets Natural Stupidity” appeared. I didn’t know then; I do now. A downloadable PDF can be found here.

Leave a Reply!

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: