Managing renewable resources via Collective Intelligence

Fabio Boschetti, CSIRO CMAR, Australia

The sustainable management of renewable resources depends crucially on understanding the feedback interaction between the human demand and the resource response, resulting in complex, often counter-intuitive dynamics. We cast the exploitation of a renewable resource within a game-theoretical framework, in which agents try to out-compete each other by predicting locations of minimum exploitation. In a previous work we showed that a Collective Intelligence (COIN) approach can achieve almost optimal resource exploitation outperforming fully competitive or fully collaborative strategies. Most important, COIN allows the team of agents to adapt to resource variability faster than other methods.

We also extend the method to more realistic scenarios. We allow agents to competitively choose what strategy to adopt and check whether COIN can co-exist with other strategies in an evolutionary stable configuration, thereby implicitly mimicking the role of defectors in the community. We find that the stable balance between fully selfish agents and COIN depends subtly on the ratio of instantaneous demand to available resource. This suggests this ratio could be used for management purposes as an indicator of the level of resource exploitation.

We also notice that COIN performance is extremely robust to uncertain information, especially when longer records of historical catch are accounted for. Finally, when agents carry out a prediction of future resource dynamics, COIN is able to find a reasonable balance between resource exploitation and resource conservation, which greatly enhances the likelihood of the resource to recover from possible crashes.