How to deplete natural resources. Decision making algorithms for geographical agent based models.
In the tragedy of the commons, you scramble to consume a resource before anyone else does.
If the resource is distributed spatially however you need to balance exploring areas where the resource might be hidden with exploiting areas you know already.
Here I present a set of algorithms to model efficiently this trade-off. I'll begin with bandit algorithms and highlight the problem of regret scaling linearly. I will then show how in agent-based models we can tame this dimensionality curse both through societal dynamics and by inferring reward structures.
Plenty of applications to fishery management follow.