Duncan J. Watts. Six Degrees: The Science of a Connected Age. New York: W.W. Norton & Company, 2003, 368 pp., $27.95 paperback.
Review written by Joel Podolny, jpodolny@wjh.harvard.edu, Harvard University
When I think back to my first exposure to the field of network analysis, I remember two ideas as being especially provocative. The first of these was Granovetter’s insight into the strength of weak ties. The second of these was Milgram’s small world study. If one were to consider the citation count to Granovetter’s study of job search or reflect on the way in which Milgram’s stylized fact of six degrees of separation has become common knowledge, one would be tempted to conclude that weak ties are where the really interesting network phenomena lie.
However, if one looks at the way that the field of network analysis has evolved over the last several decades especially in the area of organizations (the area with which I am most familiar), the dominant trend in network research has been in the direction of an almost exclusive focus on strong ties. This trend is clearest in the egocentric network survey data that researchers (including myself) have employed, but it is also implicit in the increasing volume of research on networks within a single-site firm. Studies of full networks within single-site firms invariably circumscribe the boundaries of the network so that that the numbers of individuals in the study is no greater than in the double-digits.
The reason for the exclusive focus on strong ties within a single locale seems pretty obvious – empirical tractability. With practically any focal individual being one, two, or three steps from tens of thousands of individuals, it is difficult to explore the way in which this broad, diffuse set of connections enables action and outcomes. This empirical intractability of studying weak, indirect ties has, in turn, led researchers to shine their theoretical spotlight on the thirty or so ties with which an individual is most actively engaged.
Against this backdrop appears Duncan J. Watts’ book Six Degrees: The Science of A Connected Age. This book is the latest in a series of works in which Watts – sometimes with co-authors and sometimes alone – reinvigorates interest in the small world phenomenon and in so doing raises provocative questions about the importance of those more ephemeral connections for the outcomes experienced by individuals, organizations, social movements, and society as a whole.
Unlike Watts’ impressive, highly technical first book Small Worlds, Six Degrees is written for a general audience (as evidenced by the fact that the first endorsement on the back cover is from Alan Alda). In trying to highlight the implications of the “small world” for people’s lives, he moves through a dizzying array of examples, such as the Ebola virus, financial panics, manufacturing at Toyota, and the response of New York City and various firms to the tragedy of September 11th.
When I read Watts’ first book, I was struck by his command of what could (and implicitly what could not be) concluded from the simulation models that he employed for studying the small world phenomenon. Too often the results of a simulation seem highly contingent on a particular configuration of parameters; in contrast, Watts mapped out the relevant parameter space to highlight exactly which combination of parameters would yield a small world, and in the process, showed that there are good reasons to believe that a vast array of network structures will display small world properties.
What seems most impressive about this book is Watts’ command of how features of a model need to be tailored to reflect aspects of a particular reality in which one might be interested. This book is not about a generic “small world model” that can be applied in a diverse array of contexts. Rather, this is a book about how a basic set of ideas and models need to be modified and tailored to take into account unique features of different contexts. For example, in exploring the implications of small world properties of a network for contagion, Watts draws on a general class of models from physics called percolation models. In drawing on these models, he shows a wonderful sensitivity to how the particular features of the model need to be modified if the contagion phenomenon in which one is interested involves a biological virus, a computer virus, or a technological innovation.
To be clear, Watts does not include the details of the models in this book though many of the research papers that underlie the book’s ideas are referenced in the bibliography. Rather, he presents the reader with the basic features of a model, the challenges in applying the model in a particular context, the changes to the model that he and his collaborators (and sometimes competing researchers) made to reflect the features of the context, and then the conclusions that he draws from the modified model. Researchers seeking to employ Watts’ models will obviously need to look at his more technical writings, but researchers who simply want to know what assumptions underlie his various conclusions will probably be satisfied with this level of detail.
Watts extracts various bits of practical wisdom from his and his collaborators modeling efforts. For example, he argues that to the extent that a society is indeed a small world, a needle exchange programs is an especially valuable means for fighting the diffusion of AIDS. The reason is that needles are not only shared by friends but by complete strangers, and thus they are a source of random short-cuts linking otherwise disconnected clusters of individuals. He argues that Microsoft might want to think of putting out different versions of email programs with relatively low levels of commonality/compatibility between them to reduce the degree to which its software and therefore the entire internet infrastructure becomes susceptible to a single computer virus.
In my view, some of the most provocative ideas in the book are in the area of organization design. Watts argues that the ability of organizations to adapt depends on the degree to which they can take on small world properties, thereby efficiently reducing the social distance between individuals who might need to be connected. However, the challenge in turning an organization into a small world is that there are limits on the volume of information that can effectively be processed along any particular path. Therefore, a key characteristic of adaptable small world organizations must be the capability of identifying overburdened connections reducing the information burden on those connections. Watts argues that this capability will be enhanced to the degree to which the organization possesses “multilevel connectivity”. At the highest levels in the organization, there is a high density of ties (some of which span across organization levels), and there is a decreasing density of ties as one moves down the levels in the organization.
If I had one concern about the research agenda encompassing Six Degrees, it is that Watts provides few clues for the type of systematic data analysis that might accompany his theoretical work. To be sure, there are some clever analyses of network structure referenced in this book, such as one in which a model of affiliation networks is applied to board interlock data. However, the primary empirical referents for the models are illustrations, and illustrations are not empirical tests. This comment obviously takes me back to the beginning of the essay. If network researchers have focused on the strong ties within a relatively well-defined group or around a focal individual for reasons of empirical tractability, then this book may not jolt the field out of its current focus.
However, at the end of the day, that may not be Watts’ problem, but the problem for those of us who do empirical research. Watts has reminded us that the large-scale properties of a network system matter, and it is up to us to figure out how to devise empirical studies that take these properties into account.