Reinforcement Learning is not a Causal problem

TL;DR

This paper argues that Reinforcement Learning, despite involving agent actions, fundamentally does not address causality, based on an analogy with mathematical structures and algebraic levels of information.

Contribution

It introduces an analogy between mathematical structures and algebraic information levels to demonstrate that RL is not inherently a causal problem.

Findings

Reinforcement Learning does not inherently solve causal inference.

Mathematical analogy shows RL's limitations in causal reasoning.

Current RL formulations are not aligned with causal problem frameworks.

Abstract

We use an analogy between non-isomorphic mathematical structures defined over the same set and the algebras induced by associative and causal levels of information in order to argue that Reinforcement Learning, in its current formulation, is not a causal problem, independently if the motivation behind it has to do with an agent taking actions.