Modeling Sustainable Resource Management Using Active Inference

Abstract

Active inference helps us simulate adaptive behavior and decision-making in biological and artificial agents. Building on our previous work exploring the relationship between active inference, well-being, resilience, and sustainability, we present a computational model of an agent learning sustainable resource management strategies in both static and dynamic environments. The agent’s behavior emerges from optimizing its own well-being, represented by prior preferences, subject to beliefs about environmental dynamics. In a static environment, the agent learns to consistently consume resources to satisfy its needs. In a dynamic environment where resources deplete and replenish based on the agent’s actions, the agent adapts its behavior to balance immediate needs with long-term resource availability. This demonstrates how active inference can give rise to sustainable and resilient behaviors in the face of changing environmental conditions. We discuss the implications of our model, its limitations, and suggest future directions for integrating more complex agent-environment interactions. Our work highlights active inference’s potential for understanding and shaping sustainable behaviors.

Appendix 1 Figures

Fig. 5.

Prior preference for Case 2. The agent’s preferences are changed so that, unlike case 1 and 1.1 it no longer has a preference over food left. Its only non-uniform preference is to have a preference over satiety.

Fig. 6.

Example run from Case 2 without learning enabled on the left and with learning enabled on the right, but starting with an extreme B matrix setting (probabilities set to 1 or 0, on the left three plots and middle three plots, and high but non-extreme values on the right). The agent dies quickly, just as the randomly set values of the B matrix in plot 6, and is able to learn on the right.

Fig. 7.

Case 2 with strong prior preferences on both high satiety and high food left states. On the three top left plots, the agent has no learning, and on the top right, the agent has learning. On the bottom, we can see that survival time is vastly different with and without learning, as the preferences affect the behavior of the agent.

Fig. 8.

Case 2 in a changing environment where food and satiety change at different time rates. The three top left plots show the results without learning off, and the three top right plots show the results with learning on. The bottom plot represents the comparison between the survival time over 10 time steps. Food increases at a slower rate (0.5 units per step) when not eating and decreases at a faster rate (1 unit per step) when eating. Satiety decreases faster when not eating (0.2 units per step) and increases at a different rate when eating (0.8 units per step).

Fig. 9.

Case 2 in a changing environment with seasonal simulation. The environment is built to simulate switching seasons summer/winter. With learning off the agent is not able to survive throughout the first season (top left plot and bottom plot). With learning on, it survives but then quickly dies throughout winter (top right plot).

Fig. 10.

Case 2 example runs with policy length = 1, left plot without learning, right plot with learning, and survival time on the bottom.

Table 5. Results for Case 2 extended variations