Friday, August 9, 2013

Complex Adaptive Systems
 
Ideas Seminar - Mirsad Hadzikadic
Date: August 7, 2013


Talking with Mirsad I was interested to hear that people just don’t understand the reductionist logic of simulations. They don’t like that their actions can be reduced down into a simple probability. They see themselves as complex human beings with the ability to make decisions from their own free will. However, I think they are missing the point. The study of complex systems is not about predicting the actions of each individual actor, but rather, finding the aggregate of all of the actors’ solutions. When the framing of the problem is reduced to one decision, finding the probability that people will make one choice over another is quite simple and reliable, and that’s why the study of adaptive complex systems works. As long as the system does not suffer an even that requires it to adapt, the simulation will stay relatively accurate.



I’m hesitant to say that a simulation is just a good estimation, as meteorologist have shown us, simulations of adaptive complex systems can be very accurate, but many times experience a fade in their accuracy as they try to predict further and further into the future. In the case of meteorology, the system is simulated a number of times, due to the chaotic nature of its evolution, and then these simulations are interpreted by people to give a clear understanding of the weather over the next few days. It seems that as the number of nodes in the system increase, and the interdependence of each node increases, the system become harder to predict, even though simulation is the best way to solve these types of problems. Thus, the most important part of creating accurate simulations is to frame the problem in such a way that it reduces the complexity down to a few variables that can be represented by a probability. As the interdependence of the variables increases, so does the complexity of the system, and as such the accuracy of the simulation is diminished.



In the end, it is not really about the nodes, or the people that they represent. For a node-centered simulation to be accurate we would need to develop a bottom up approach. While AI is certainly developing quickly, we still don’t have the computational power to compute such a system where the movements of each node are predicted. Thus, the simulations present in the study of complex adaptive systems seek to predict the outcome of a system without having to worry about the accuracy of the system at the node level.

No comments: