A Coalition to Protect Research Congressional Briefing

Six Degrees of Separation:

Using Social Network Research to Inform

Public Health and National Security

On June 10, the Coalition to Protect Research (CPR), along with COSSA and 12 other organizations, cosponsored a Congressional Briefing on the importance of social network research for public health and national security policy.  The Coalition to Protect Research is a coalition of 60 organizations committed to promoting public health through research. CPR is headed by Angela Sharpe, Deputy Director for Health Policy of COSSA, and Karen Studwell,  Senior Legislative and Federal Affairs Officer for the American Psychological Association (APA).

Katherine Stovel, U.S. Major Brian Reed, Duncan Watts, and Sally Hillsman

Social networks are based on patterns of interaction among individuals, organizations, groups, and even countries. Each individual’s unique social networks can have profound effects on their physical and mental health as well as their personal safety. 

Scientists working with social network models have identified how infectious diseases like SARS and STDs are transmitted across communities of individuals.  Similarly, researchers working with the military have employed social network analysis to identify and track terrorist networks as well as locate terrorist targets.   Briefing attendees heard from four distinguished scientists who are applying social network analysis to critical issues ranging from high-risk adolescent behaviors to military intelligence.

U.S. Major Brian Reed from the Center for Research on Military Organization at the Department of Sociology, University of Maryland College Park discussed social network analysis and resistance networks.  Katherine Stovel, a sociologist from the University of Wisconsin, spoke about romantic networks among adolescents.  Duncan Watts, a professor of sociology Columbia University and author of the book Six Degrees: the Science of a Connected Age (W.W. Norton, 2003), talked about the unpredictability of epidemics, and American Sociological Association (ASA) Executive Officer Sally T. Hillsman moderated the session. 

 Hillsman opened the briefing by pointing out that social network analysis is often seen by non-academics as “soporific” and that in the past, it has rarely left the seclusion of academe.  Now, she said, this area of study is becoming more visible outside of the academic world, emerging in sometimes unexpected places.  Hillsman also gave reference to the play by John Guare that inspired the title of the seminar, “Six Degrees of Separation,” explaining the theory that any two people in the world are connected by a network of six or fewer people.  The practical implementation of social network analysis, she said, can lead us to find “meaningful simplicity in the midst of complexity.”

Resistance:  How Do We Depict a Network

 That Does Not Want To Be Found?

Reed elaborated on the application of social network analysis to military operations.  He explained that the role of such analysis is to somehow infer organization in the absence of visible structure, which has played a pivotal role in attempting to track down terrorist, or “resistance” networks.  As Reed pointed out, “How does one depict such a structure if the group does not want to be known?”  He named the Bolsheviks and Viet Cong as prime historical examples of “cell networks,” the study of which may prove useful in depicting the structure of similar networks in Iraq or Afghanistan.  Being able to diagram these networks, in concert with terrain analysis and an understanding of diverse populations, enable intelligence preparation in the battlefield, or (IPB).  These factors are often studied in order to choose locations for and timing of maneuvers or attacks. 

Reed presented three main ideas for the audience to take home.  First, resistance networks do not behave like normal social networks.  For example, strong ties between people can appear to be weak, incomplete results can compromise the network picture, it is often ambiguous who in the network is “in” or “out,” and the dynamic of the group is constantly changing.  Second, he explained that the best practical application of social network analysis may be to identify suspects and then map their networks to see where they may lead.  Third, he argued that social network analysis can allow for better prediction of certain future behaviors, making for clearer evidence and a better likelihood of prosecution.  Once a network has been identified, we are able to investigate its structure, how it works, how it is connected, and how to best destabilize it.

Stovel: We Must “Rethink” Individual Risk

Within Sexual Networks

Stovel shifted the discussion to the National Longitudinal Study of Adolescent Health, or the “Add Health Study,” an ongoing study on adolescents attending 143 schools in 80 communities throughout the U.S. and sponsored by 17 federal agencies.  It was designed to measure the social contexts of adolescent health.  The study includes in-school questionnaires, in-home interviews, and parent questionnaires.  Using wave one data, she talked about a segment of the study that examined the romantic networks among youth.  In mapping these networks, the data indicated several network structures, including “romantic cascades,” or networks where high and low-risk people interact without cycles (within which the spread of disease is slow, but where subtle changes in romantic chains can cause massive effects on disease risk).  Stovel illustrated her points by using maps of the networks in one of the study’s high schools, where participation was 100 percent.  Sheemphasized that we must begin to “rethink” individual risk; “risk” is not necessarily what individuals do, but also who they do it with and the type of network in which they are embedded.   

Conventional Epidemic Models

May Be Flawed

Watts began his presentation about the unpredictability of epidemics by explaining why many of the longstanding conventional models used to predict epidemics are fundamentally flawed.  The SIR (Susceptible, Infected, Recovered – or Removed) model, for example, assumes that the “susceptible” and the “infected” run into each other in random mass action.  Watts elaborated on the reasons that so many models use this “mass action” assumption: “… the standard model is to assume that individuals bump into each other at random.  And so the probability of an ‘infected’ running into a ‘susceptible’ is just proportional to the product of their population sizes… this is a tremendously simplifying assumption because now you don’t need to model the disease at the level of individuals; you can instead model it at the level of populations.”  He added: “… this is a very powerful assumption.  It enables you to write down differential equations instead of having to do very complicated simulations.”  Watts argued that this is not accurate because the nature of epidemics is such that they either infect very few people, eventually disappearing, or the disease spreads rapidly and a large chunk of the population becomes infected. 

He also challenged the notion that epidemics only “peak” once during their run.  The reality, he said, is that due to their introduction into different parts of the world with sometimes vastly different living conditions, any given epidemic can peak several times.  Diseases will often circulate in a certain locale, and then “jump” to a distant place because of a few individuals who travel.  “And furthermore,” he added, “temporarily there is a lot of uncertainty as well.  Just because the disease is burning itself out doesn’t mean that it can’t find itself a new population and start right back up again.” According to Watts, these examples serve as illustrations of the need to study population structure in order to better predict and understand epidemics. 

During the question and answer period, Mary Jo Hoeksema of the Population Association of America (PAA) asked about the role that technology has played in both the rise of resistance networks and their identification.  Reed responded that new technology has made communication more vulnerable on both sides, but has also sped up the communication process.  For this reason, he pointed out, we are starting to see these networks revert back to more primitive, secure modes of communication. 

Implications of Social Network Research

for Public Policy

Hillsman interjected a question of her own, asking each presenter whether they think that data on social networks are improving and if there are actions that government and policymakers can take in order to further the science.  Stovel responded by saying that data collection efforts are improving, but that more understanding of network patterns is needed.  She observed that at some point, it would be useful to integrate mathematical modeling into social network analysis in order to see if there are some identifiable patterns emerging.  Watts added that “there is a tremendous advantage going on now in recording who interacts with whom and possibly even with what consequences.  Unfortunately, for publicly funded researchers, that is almost all in the corporate world and it’s proprietary.”  In many cases, he argued, progress is encouraging and simply needs more data, while other areas are not as promising as the others and may need a new approach altogether. 

David McMillen of the House Government Reform Committee questioned the extent to which social network analysis can be ethically and justly used in the public policy arena, for example, those who have distant relations or one-time interactions with al Qaeda members being hunted down as terrorists; “… how do we draw that line between what is valuable research and when is it appropriate to move that research into the policy realm where the consequences are quite big?”  Watts answered the question by saying: “The problem with individuals is that they are very complicated and two people who have very similar histories and backgrounds can end up behaving very differently… even if you happen to be genetically identical you can have very different life courses depending on any number of variables… So you have to ask the right questions and unfortunately, a lot of the questions that people want the answers to, particularly in the policy world are the wrong kinds of questions scientifically.”

Stovel also tackled a question about using populations in other countries to study the progression and network pattern in the U.S. AIDS epidemic.  She explained that currently, mathematicians and workers in the field are collaborating in places like Southern Africa to come up with better models of disease transmission.  “… I think this is a context where we really do need people working with different kinds of expertise to feed better estimates of social behavior, sexual behavior, whether it’s sexual relationships per se or other kinds of things that might bring people into proximity with one another…”

The final question of the day came from Barbara Solt of the Institute for the Advancement of Social Work Research, who asked about the need for getting data on networks that are not terribly popular, such as disease transmission among truck stop prostitutes – one of the 150 sexual health research grants funded by the NIH that have been the subject of congressional inquiries, the most recent at a House Energy and Commerce hearing on the possible reauthorization of NIH (see UPDATE, March 21, 2005).  Stovel responded “… I think the more we study networks, the more we find both that there are strong similarities; there are some principles that seem to guide people’s interactions with one another…  And yet there is important variation as well.  And to the extent that we can understand that variation… we then will be able to begin to think about the consequences of networks in a more robust way.” 

Concluding that there are “no easy answers here,” Watts emphasized that as with the prediction power of any variable, social network theory cannot yet provide definitive answers to terrorism or disease because it is still in its fledgling years of study.  He drew an analogy between the future of social network analysis and the future of DNA back in 1953; years of research and billions of dollars have been put in to harness the power of this science.  Watts concluded by saying “this is science like any other kind of science.  And if you want answers, you have to pay for them… you don’t get there with just one person sitting and dreaming up his or her theory of the world in their office... You need to industrialize these things.”

The organizations cosponsoring this briefing included: American Academy of Political and Social Science, American Psychological Association, American Sociological Association, Association of American Medical Colleges, Association of Population Centers, COSSA, Federation of Behavioral, Psychological and Cognitive Sciences, Institute for the Advancement of Social Work Research, Population Association of America, Reproductive Health Technologies Project, Society for Research in Child Development, The AIDS Institute, and The Mautner Project, the National Lesbian Health Organization. 

Updated June 16, 2005

"