Catching the “Network Science” Bug

Recent efforts to develop a universal view of complex networks have created both excitement and confusion about the way in which knowledge of network structure can be used to understand, control, or design system behavior. This paper offers perspective on the emerging field of “network science” in three ways. First, it
briefly summarizes the origins, methodological approaches, and most celebrated contributions within this
increasingly popular field. Second, it contrasts the predominant perspective in the network science literature
(that abstracts away domain-specific function and instead focuses on graph theoretic measures of system
structure and dynamics) with that of engineers and practitioners of decision science (who emphasize the
importance of network performance, constraints, and tradeoffs). Third, it proposes optimization-based reverse
engineering to address some important open questions within network science from an operations research
perspective. We advocate for increased, yet cautious, participation in this field by operations researchers.

Through the company, Innovative Scheduling, that I founded a few years ago, I have been involved in developing network models for solving some very complex and large-scale decision problems arising in logistics and transportation. These problems are very large-scale mixed integer programming problems containing billions of decision variables. Network models allowed us to simply the decision problem and make them tractable. Selling these models to company executives and persuading them to use them has also been a part of my job. I have found that networks offer a very intuitive way to explain methodology and convince them the power of modeling. Networks can be grasped by a common man. They are very intuitive and so are the algorithms to solve network models. Thanks to the intuitive aspect of network models, I have been fairly successful in explaining complex network algorithms to executives who have had no degree in operations research. I owe our success to network models; I have thus found network models to be highly rewarding.

Research activities [...] that cut across intellectual traditions produce
the most innovative ideas. March (1991) characterized the tension between the two activities as a
trade-off between “exploiting” existing knowledge stocks as opposed to “exploring” new knowledge
domains. Since our business rewards innovation, intellectual arbitrage is a necessity. a significant
barrier to arbitrage is making sense of ideas and innovations that lie outside our immediate area of
expertise. Even if we were all inclined to work across intellectual domains, the activity is costly. Any
critical review that increases our understanding of a significant development outside our field and
puts those developments in a familiar language provides a valuable service. Each reader must decide
if Alderson has crossed that threshold. He has for me and he has also identifi ed a fundamental
concern with the new science of networks. In particular, analysis in the new science has focused on
nodes or relationships). Alderson worries the “content-free” approach could bias our understanding
of network dynamics. I agree. Our challenge, therefore, is figuring out what content can be set aside
while we focus on more general processes and what content cannot be ignored because it defines
the substance of our theoretical enterprise.

I appreciate the comments of Ray Reagans on the implications of this article for organizational behavior, which is outside my area of expertise. Much like OR, the analysis of social networks predates recent interest in Network Science by several decades.

I once began a decision analysis tutorial by saying, “Hi, I’m Lee, and I’m an optimizer. I haven’t
optimized in six months.” That was a long time ago, but the picture hasn’t changed much. I have
always been drawn to optimization, and did my dissertation research in NLP under Fiacco and
McCormick. I still fancy myself an optimizer, but in all honesty I fi nd myself doing a lot more
DA, simulation, and, of course, managing. So I was pleased to have the opportunity to read David
Alderson’s article on network science. Honestly, I hadn’t yet witnessed the bifurcation of network-
based analysis into network science and traditional OR network optimization, but Dr. Alderson
makes a compelling case.

— Given the plethora of books and scientific articles with networks published in the OR literature,
why is it that only fairly recently that such contributions are being cited/acknowledged by those
working on networks in physics?

— As Alderson notes, his essay did not directly discuss computer science and that discipline’s perspective on network science. As is well-known, computer scientists publish primarily in proceedings,
rather than journals, and their representation in terms of refereed journal articles is hard to track
(see Alderson’s Table 2 footnote). Are we sufficiently communicating with this community?

— Given the recent substantial growth in funding for network science (and engineering), how do we
as a community present ourselves as leaders in this area (or because of time delays — followers)?

— More broadly, should we be involved in better popularizing our results and, if so, are we doing
everything possible in this dimension as a community?

There are many more outlets in physics than OR (as well as there being more researchers trained in physics than OR), and the turnaround times from submission to publication are much shorter. This undoubtedly results in a greater volume of publications in physics than OR.

…

But a disadvantage in terms of the quantity and frequency of publications does not relieve the OR community from its need to be proactive in the debate about complex systems of practical importance. Currently, this debate centers on Network Science, and decision makers ranging from research sponsors (DoD, NSF, DOE, NIH), to corporate executives, to government authorities (e.g., DHS) are paying attention.