An Economics With Verbs, Not Just Nouns

So what is Arthur's alternative? As an economic theorist with extensive mathematical training, he suggests that economists open themselves up to an alternative kinds of math--the mathematics of algorithms. He writes: 

The reason algorithms handle processes well is because each individual instruction, each step, can signal an action or process. Also, and here is where process enters par excellence, they allow if-then conditions. If process R has been executed 100 times, then execute process L; if not, then execute process H. Algorithms can contain processes that call or trigger other processes, inhibit other processes, are nested within processes, indeed create other processes. And so they provide a natural language for processes, much as algebra provides a natural language for noun-quantities. Frequently algorithms include equations, and so sometimes we can think of algorithmic systems as equation-based ones with if-then conditions. As such, algorithmic systems generalize equation-based ones, and they give us a new mode, a new language of expression in economics, although one that may look different from what we’re used to. ...

The world revealed here is not one of rational perfection, nor is it mechanistic. If anything it looks distinctly biological. Its agents are constantly acting and reacting within a situation—an “ecology” if you like—brought about by other agents acting and reacting. Algorithmic expression allows novel, unthought of behaviors, novel formations, structural change from within—it allows creation. It gives us a world alive, constantly creating and re-creating itself.

Arthur and others who work in "complexity economics" have been creating and working with these kinds of models for decades. As Arthur writes in this paper, such models can be viewed as simulations" or "laboratory experiments"--that is, given certain starting points and behavioral rules, if you allow the algorithm to evolve many times, what kinds of outcomes are more or less likely to evolve? 

All this is fair enough. My own sense is that algorithmic methods can be especially useful in showing how seemingly mild and plausible rules can sometimes lead to unexpected and even disastrous outcomes, and how small changes in initial conditions or in underlying assumptions about behavior can lead to dramatic differences in how outcomes evolve. 

View single page >> |
How did you like this article? Let us know so we can better customize your reading experience.


Leave a comment to automatically be entered into our contest to win a free Echo Show.