This interesting article in the Economist shows how American politics is becoming increasingly polarized partially because when people move, they locate in areas with similar cultural preferences – be it granola or shotguns. When I lived in the States, I was always fascinated by the difference between Vermont (Birkenstock and yogurt country) and New Hampshire (main business: roadside hubcap emporiums). As it turns out, this split between liberal and conservative is happening all over the country, and you end up with the curious situation where the United States from the melting pot evolves into a salad, with rather few ingredients.
All this is interesting, but hardly relevant for technology, no? As a matter of fact, not: I am currently working on a research project with nGenera, called BST: Putting Business Simulation Technologies to Work. Simulation allows us to see the aggregate effect of many small decisions.
One of the early books showing the importance of this is Mitchell Resnick‘s Turtles, Termites and Traffic Jams. In this book, Resnick demonstrates a number of simulations programmed in StarLogo (a parallel version of Logo, a programming language originally created for children.)
One simulation in particular (caveat: this is from memory, my numbers may be wrong here) is pertinent to the polarization of America: The effect of weak preferences on clustering. Resnick constructs a 100 x 100 matrix where each cell is inhabited by either a black or white dot. Each dot can “think” (i.e., have preferences) for itself, and the simple preference each dot has is the unless it is living in a neighborhood with at least two of its own kind (“neighborhood” defined as the 8 cells sharing a side its own cell) it will move, randomly, to somewhere else. Note that this is not a strong preference: A dot of one kind will happily inhabit a cell where 6 of its neighbors are different, as long as two are the same. (A more thorough description, with images, is here.)
In a surprisingly short time, the initially well distributed matrix transforms into clear clusters (even bands) of white or black. Importantly, this process, when viewed from a distance, seem to be conscious, yet the relatively mild preference exhibited by each individual dot seems rather harmless. It may be tempting to ascribe the segregation to some conscious plot, failed policy or other single cause. It is a very powerful demonstration of the aggregate and cumulative effect of small decisions and weak preferences – and simulation is the only way to make it apparent.
Resnick’s book shows similar uses of simulations to understand ant foraging strategies and traffic jam formation – and some of the insights have been put into use in real life. For instance, traffic lights at on-ramps that introduces cars into traffic flow in an even stream rather than random groups is, as far as I know, a direct result of simulations of traffic jam formation.
In science, business and politics, we are moving from isolating single factors and varying them to understanding interaction patterns between many small components. Simulation allows us to understand this – the challenge lies in understanding where and how this very powerful tool can give insights.
And there you have Vermont and New Hampshire, Virginia and Maryland: The results of weak preferences over time. Perhaps we could simulate some real political discussion at some point?