Endogenous Growth in a Swarm Economy: Fighting Time, Space, and Complexity

Agent-based methods have provided economics with new ways of modeling economic processes. We have found these possibilities useful with respect to modeling endogenous growth, in particular the modeling of learning agents that must cope with real world growth constraints such as time, space, and complexity. Time and space are immediately given once a two-dimensional space with agents moving in time is introduced. Complexity is modeled by using single-cell neural nets as a metaphor for production units and XOR-problems as metaphors for demand signals. The production units learn by changing the weights in their neural nets, but their structure is too simple to learn to respond correctly to demand signals. However, as entrepreneurs, production units may connect more production units into more complex neural nets and thus improve the efficiency. We thus have an artificial economy where economic growth is driven by demand but limited by both demand and supply sides. On the supply side, neural nets must receive demand signals to learn and thus become more efficient. Besides time and space constraints in the search for efficient production, the demand side is constrained by the distribution of the generated wealth since agents have limited access to credit. An interesting feature of the model is how it produces cycles around the growth path.