Social networks: Guilt by association?

Our social networks tend to show distinct social circles

One of the most interesting stories that has emerged from the NSA surveillance scandal is, in my opinion, the revelation that the intelligence agency has been using some of the data collected to conduct social network analysis (SNA) on the general population.

This is a topic that is discussed in some detail in my book. Social networks are groups of individuals that are connected to and interdependent from one another. These individuals have interactions that can range from family ties to friendship, employment information and development. A social network can be understood as any bounded set of connected social units. Social networks then rely on three key building blocks: the boundary of social elements studied, be it a family, a tribe, or a country; then the connected element between the social units, which are the links that tie the units together; and the definition of social unit itself, these are usually individuals, but also can be groups of groups, so we could have social units consisting of organisations and institutions. SNA is therefore a systematic way of looking at these networks in an analytical fashion by using graph theory in order to provide useful information about the group.

The NSA seems to be using collected metadata from telephones and Internet services to try to draw an accurate social map of not only the USA, but of the world. This may seem like an innocuous exercise, but it is possible to learn quite a lot from something so seemingly innocent as the information gathered from your communications. One of the most difficult aspects of SNA is to try to define a network, and our communications in any medium act as nifty delimiting parameters with which one can draw an accurate picture of a person’s social environment and circumstances. There are tools available that can help you get an idea of the importance of metadata for SNA, such as Project Immersion, which analyses your social network through Gmail. Trying Immersion is an eye-opener when it comes to mapping one’s communications, and I personally found its accuracy daunting, it was interesting to see my social network shift before my eyes as I moved the time scroll bar.

It would be easy to dismiss such tools, they can help to draw pretty pictures about one’s Linkedin connections or Facebook friends, but why is it of interest to the NSA? Despite the attempts to dehumanise them, terrorists are social entities as well, they have friends, family, acquaintances, and more importantly, they also tend to move in terrorist networks. The Lone Wolf is actually a rare occurrence, and most terrorists tend to operate as part of a group. The difficulty in catching Bin Laden was precisely that he carefully excised himself from social interactions, and went into almost complete isolation. The fact that one of his few connections was used to discern his location is one of the greatest successes of social network analysis. So, the NSA has been using metadata gathered through their snooping programmes to try to identify terrorist networks. Let’s say the NSA has identified a suspect, by analysing his connections, it might be possible to get a good picture of possible members of a group, once other networks are identified and filtered. See for example this representation of the September 11 terrorist network (from here):

In fact, this produces quite accurate pictures of possible leaders, and in reality it can be used to identify members and other possible suspects that might warrant further examination.

So, is the NSA programme justified? Not necessarily. While social network analysis can prove to be useful in identifying possible terrorist networks, its deployment in the general population clearly raises the spectre of false positives, and even of guilt by association. Do we really want to live in a world where casual connections to someone who might have done something wrong is enough to unleash further surveillance? There are serious privacy implications to this approach. Moreover, any change to the definition of “terrorist” could be used to justify greater surveillance.

Unfortunately, some sectors of the press have been latching on this subject from the wrong perspective. In this piece, the author raises seems to mistakenly read two independent pieces of research to come to the conclusion that large numbers may already be terrorist suspects. The author cites a study conducted some years ago which found that the Internet displays 3.3 degrees of separation, that is, the average social distance between people online is of just over 3 connections. That means that you can get to anyone in just over 3 hops. As the NSA is using 3 degrees in its SNA analysis, the conclusion is that you might already be a suspect as chances are you might be already connected to a terrorist.

This is slightly sensationalist reading of what is going on. Firstly, Internet connections are more like 4-6 or over degrees of separation, depending on the tool that you use (Facebook produces 4.7 degrees). Secondly, it is very unlikely that any random connection to a terror suspect will produce a positive hit. One does not only need to be a random connection to elicit suspicion, one has to be a part of a network. Take my very own Linkedin connection map:

This shows several distinct groups, with independent and loosely connected individuals peripherally attached in grey. Imagine two terrorist suspects were found in the large purple group on the right, this might create some suspicion, and it could warrant further investigation. But if you are loosely connected to a person, but clearly does not involved a group, then it is possible that no action will be taken.

Do not get me wrong, I still believe strongly that NSA metadata gathering and its surveillance programmes are dangerous. Indiscriminate data gathering cannot possibly be of any use in and of itself, and I cannot fathom any justification for drawing a picture of the social networks present in the general population. Without intelligence, those networks simply represent pretty pictures with no context. Such indiscriminate surveillance increases the risk of being found guilty by association. Careful oversight and care is warranted, and from what I have been reading about the NSA and FISA, there is a serious lack of transparency.

Having said that, social network analysis can prove useful when used judiciously. My fear is that the intelligence services are extending their reach and creating truly intrusive tools that erode our rights.