Breadth and depth of search on a technology landscape
Karén Hovhannisian, Visiting Researcher MERIT, the Netherlands
This paper presents a simulation model of adaptive local search on a technology landscape, enriched by the notion of agents able to search both deep and wide. The breadth of search is a horizontal measure of the number of directions for search activities considered for at each level, while the depth of search is a vertical measure of the agents’ foresight. Thus stated the model is deducible both to the original NK model of local search where depth=breadth=1, as well as the models of perfect foresight where depth=breadth=N. The results show that while more complex technological landscapes (with the complexity measured by the level of intra-firm externalities between the elements of a given technology) require higher depth/breadth values to guarantee agents reaching a global maximum, even for maximally complex landscapes, perfect foresight is not a necessary condition for that. Further, introducing competition between the agents in the model, we observe how agents with less “insight” can ultimately outcompete more “insightful” agents, due to their higher speed of adjustment.
Date: 29 April-00 0000
