FAQ

1. Does PyCFRL work for online reinforcement learning?

No. the sequential data preprocessing method provided by PyCFRL is designed for the offline reinforcement learning setting where a pre-collected offline trajectory is available. It does not apply to the online setting in general.

Indeed, to our current knowledge, there has yet been a method specifically designed for ensuring counterfactual fairness in online reinforcement learning. Once such a method is developed, we will try to incorporate it into PyCFRL.