Symbolic Systems 205 – Systems: Theory, Science, Metaphor

Winter 2002-2003

Stanford University

Instructor: Todd Davies, Symbolic Systems

Commentary: Discovering Artificial Economics: How Agents Learn and Economies Evolve

by David Batten

commentary written by Sara Wampler

Most of the selection from Discovering Artificial Economics that we read for our seminar revolved around the differences between static and dynamic methods of modeling economic development, as well as a discussion of how these models can be applied to a historical understanding of the rise of Chicago as the dominant gateway to the American West. A smaller point was mentioned within the reading but was later overlooked in our analysis: the importance of chance on the formation and maintenance of networks. Models are important because they allow for instructive simulations, but they cannot accurately predict those chance occurrences that can cause an explosion of growth or a devastating loss. Earthquakes, floods, and hurricanes are in some ways chance events, but in many cases they are not a complete surprise; cities built on fault lines or low-lying areas can expect a natural disaster at some point, even if the specifics of time and duration cannot be preordained. However, the unpredictable events that seem to have had the most influence on the development of cities like Chicago were entirely manmade, including war, the building of transportation lines, and the beliefs of key individuals.

From Andrew Waterman’s overview of this book, it appears that Batten attempts to address issues of chance through explorations of positive and negative feedback loops in a system that he calls “coevolutionary learning.” This theory seems intuitive: people base their beliefs and behaviors on their situation and act accordingly, thus changing the situation and creating a new situation to which they must adapt. Models that explore this interplay between agents and states would seem to be more applicable to real-world situations than models that ignore the role of agents, which is why Batten is such a proponent of artificial economics. However, Batten’s artificial economics illustrates a weakness suffered by many other attempts to predict future behaviors of human groups through network theory. Whether you use the terms “sheep” and “explorers” or “mavens” and “salespeople”, no model can hope to provide accurate forecasts because models cannot really account for the proportion of “sheep” to “explorers” in a population, nor can models predict what innovations these agents will develop. Network theory can provide interesting and sometimes quite compelling explanations for the past; Batten’s book is one of several examples. But, as Batten himself pointed out, it would have been extremely difficult to determine in 1810 or even in 1830 that Chicago would become the leading western metropolis. In my opinion, this is because network theory can never understand and predict human, or even natural, ingenuity. Network theory can perhaps provide a useful heuristic for making educated guesses about the future, but because of the chance developments of human talent and ambition, network models are necessarily limited in their scope and functionality.