A Guiding Principle for Causal Decision Problems
This work addresses decision-making challenges in AI by integrating causal reasoning, though it appears incremental as it builds on existing causal and reinforcement learning frameworks.
The paper tackles the problem of decision-making under uncertainty by defining Causal Decision Problems and proposing a solution based on Pearl's Do-Calculus and Expected Utility, showing similar performance to classic Reinforcement Learning algorithms while enabling causal model learning.
We define a Causal Decision Problem as a Decision Problem where the available actions, the family of uncertain events and the set of outcomes are related through the variables of a Causal Graphical Model $\mathcal{G}$. A solution criteria based on Pearl's Do-Calculus and the Expected Utility criteria for rational preferences is proposed. The implementation of this criteria leads to an on-line decision making procedure that has been shown to have similar performance to classic Reinforcement Learning algorithms while allowing for a causal model of an environment to be learned. Thus, we aim to provide the theoretical guarantees of the usefulness and optimality of a decision making procedure based on causal information.