Friday, July 14, 2006

Will we ever be able to predict what social systems and networks will do? Perhaps globally, but likely not locally

There is a lot of interest in social systems and networks, and the use of network theory to help explain how and why social systems work the way they do. While research has shown, for instance, how different rulesets lead to various decisions or how network topology helps identify how disease spreads, one needs to keep in mind that there is a difference between local environments and global environments. What I mean by this is that the analyses done in these areas of study essentially look at results that affect the system more globally. It is quite another thing to see what happens to individual agents, since in complex systems the rules that govern individuals can be and typically are very different from the rules that govern collective behavior.

In a physical system this is similar to studying gases. We can in principle use Newton's laws to predict what should happen to individual atoms and molecules, but collectively we need to resort to a statistical/probabilistic approach. Collectively, there are set probability distribution functions for something like molecular speed, but that is meaningless to an individual molecule of the gas. In social systems, we are dealing with complex, unpredictable individual agents that make up the system, and this makes things considerably more difficult to analyze than a gas, whose individual agents are governed by deterministic rules (at least to a good approximation using classical physics). It will be quite difficult to accurately model emotion and religious fanaticism, for example, for individuals in a social system. We can guess and try to take a statistical approach, but this leaves some degree of uncertainty in results and predictions. It will be very difficult to model and predict what is going on in the head of a leader such as Osama bin Laden; there is a good deal we can only guess at, even though there has been research and progress in figuring out how his larger terror network operates and is structured. This is the difference between local and global environments and rulesets.

Whether it is trying to figure out economies, decision-making, trade networks, terrorist organizations, or worldwide transportation networks and systems, we will likely always be in a better position to understand the global structures and behaviors of those networks compared to what happens to individuals and in local segments of the larger network. There will be a 'fuzzy' area where the transition takes place between where the local environment ends and the global begins, and vice versa, similar to the fuzzy area between where relativistic effects are significant compared to Newtonian predictions, or where the boundary is between a classical and quantum systems. These are areas of study that have no clear-cut borders, and much of the answer depends on what level of sensitivity and precision you are interested in. So it goes, too, for social systems, and those who are interested in such areas of study need to keep this concept in mind. As in physical science, the difficult part will be to determine how large the error bars are on results and predictions, and will be based on how sensitive the global systems are to local perturbations caused by individuals within the system.

3 comments:

mark said...

Linked ! I'll have something up later tonight....

Daniel S. said...

Interesting that you wrote on this topic. I was actually working on social modelling when I interned at los alamos natl lab this past 10 weeks. We ended up using "framing" and "messaging" to model social interaction. I was mostly doing GUI work though, so my knowledge of our model is limited to only the technical details and not the social theory aspects.

Mark Vondracek said...

Hi daral,

How did you like Los Alamos? Next time you're in town stop in and we can chat. There is a lot of very cool research being done on these topics. Cheers!