Logo IMG


E Pluribus Unum

Brian Hayes


StarLogo is one of the most beautiful computer toys I've ever encountered. It is ingenious in concept and brilliantly executed. Writing StarLogo programs is a delight. The language offers high leverage: A few lines of code can yield deep results. The implementations work smoothly and reliably. There's an active community of StarLogo enthusiasts. The sample projects distributed with the software cover an amazing range of topics.

But StarLogo is not a Model of Everything. One limitation is that it's strictly two-dimensional. These turtles crawl; they don't swim or dive. The programmer also has no choice about the geometry or the boundary conditions of the StarLogo world. On the screen it is a square or rectangular place; topologically it is a torus, where right meets left and up meets down.

Another constraint on StarLogo models is the emphasis on local interactions. StarLogo lends itself most naturally to situations where signals pass between nearest neighbors, as in the slime mold model. In effect, the turtles have acute senses of touch and smell, but their hearing and vision are weak. Long-range interactions and global knowledge are hard to fit into StarLogo programs. This may partly explain the breakdown of the race-relations model. In a real city, residents know more about racial distribution than just the identity of their immediate neighbors, and someone seeking a homogeneous environment would not have to choose a new home at random.

Of course Resnick makes no claim that StarLogo is the universal key to understanding the world, but he does suggest some analogies and connections that may not stand up to scrutiny. In his book and other writings he speaks of the "decentralized mindset." Centralized leadership and control are being replaced by self-organizing structures and systems, he argues. He cites among his examples the collapse of centrally planned economies in Eastern Europe and the triumph of free-market capitalism. He also mentions a shift in corporate organization away from a hierarchical chain of command toward decentralized management, and the trend in computing to replace central computers with distributed processing. It's easy to see a resemblance between these developments and the kinds of models constructed with StarLogo, but I'm not convinced the connection runs very deep. In particular, economic activity is not constrained to local interactions, which makes StarLogo models of markets and businesses seem rather artificial. On the other hand, a weakness of some other simulation systems is that they totally ignore spatial factors, so perhaps this is a needed correction. In any event, StarLogo is a magnificent tool for thinking about decentralized systems.

comments powered by Disqus


Of Possible Interest

Feature Article: In Defense of Pure Mathematics

Spotlight: First Person: Jim Smith

Computing Science: Clarity in Climate Modeling

Subscribe to American Scientist