Causal Reasoning from Meta-reinforcement Learning
This work addresses the problem of causal reasoning for AI agents, offering incremental insights by demonstrating that such reasoning can arise from model-free reinforcement learning, potentially benefiting complex settings.
The paper tackled the challenge of enabling intelligent agents to discover and exploit causal structure in their environment by exploring whether causal reasoning can emerge via meta-reinforcement learning. The result showed that a trained agent could perform causal reasoning in novel situations, including selecting interventions, drawing inferences, and making counterfactual predictions, though no concrete numbers were provided.
Discovering and exploiting the causal structure in the environment is a crucial challenge for intelligent agents. Here we explore whether causal reasoning can emerge via meta-reinforcement learning. We train a recurrent network with model-free reinforcement learning to solve a range of problems that each contain causal structure. We find that the trained agent can perform causal reasoning in novel situations in order to obtain rewards. The agent can select informative interventions, draw causal inferences from observational data, and make counterfactual predictions. Although established formal causal reasoning algorithms also exist, in this paper we show that such reasoning can arise from model-free reinforcement learning, and suggest that causal reasoning in complex settings may benefit from the more end-to-end learning-based approaches presented here. This work also offers new strategies for structured exploration in reinforcement learning, by providing agents with the ability to perform -- and interpret -- experiments.