The Agent-Based Paradigm
I have been involved in several discussions regarding the usefulness of agent-based modeling to skeptics. Somewhere along the way, the discussion has always involved some version of, “it’s very interesting, but so what?” After tempering initial defensiveness, I’ve tried to highlight the advantages of the approach. The most practical is the availability of complete data on the system being examined. Within the parameters provided, the agent-based modeler has as much information as they can dream up: cross-sectional, time-series, micro-data, macro-data, etc. Additionally, agent-based models speed up time and can, therefore, provide insight into the development of the world that we may have missed otherwise, these models can tell us where to look.
However, after much thought, I’ve decided that the primary advantage is to change the way you think, it did for me anyway. It was a paradigm shift. Perhaps this is too dramatic, but the world is agent-based. The issue is that frequently, mindsets tend to be static. By this, I mean that we are all runnin’ regressions or looking for equilibrium in a closed-form model. We have data on this, that, or another thing, and we connect the dots then measure the slope. Even time-series data are motionless. Phrased differently, conventional analysis frequently feels lifeless.
Agent-based models are alive. The modeler sets up the rules and lets it run. Watching the output and imagining how the agents are acting to produce the observed output. What are these little guys doing in there? Does this output make sense? The dynamism is exhilarating. The process is alive and, therefore, so are your (my) thoughts.
Brian Arthur, quoting Robert Venturi, highlights the term “messy vitality”. The world has a deep richness to it, a complexity, a messy vitality. It is alive and is better examined by feel. The world, my world at least, has a messy vitality and so too should our models. Mine should, at least.