SAN-RL: combining spreading activation networks and reinforcement learning to learn configurable behaviorsThis approach provides the agent with a causal structure, the spreading activation network, relating goals to the actions that can achieve those goals. This enables the agent to select actions relative to the goal priorities.