InventSim: An Agent-Based Model of Product Invention

Authors

Abstract

This study develops an agent-based simulation model (InventSim) of the process of product invention. Invention is conceptualised as a search on a landscape of product design possibilities, by a population of payo_-seeking agents (inventors). The agents employ a set of search heuristics in searching this landscape which embed individual & social learning, and expectations of future payo_s to proposed product inventions. In the simulation, agents compete for payo_s and co-evolve based on each other's actions. The simulation experiments examine the impact of agent's search heuristics on the societal rate of inventive progress. The results indicate that the forward-looking heuristics of election and thought-experiments are crucial in driving forward the inventive process, even when payoff expectations for proposed product inventions are noisy.

Journal

Inteligencia Artificial, Revista Iberoamericana de I.A., Vol. 9, #28, pp. 77-84, March 2005