Thm. 3.2] we can construct an SCM

RCM. Finally, we illustrate the power of this conciliatory perspective by pinpointing

We now derive an estimator for the gradient of the square-root of the variance (i.e., standard deviation) of

Importantly, our results also show that subpopulation-based recourse is the right approach to adopt when assumptions such as additive noise do not hold.

Pearl [2009] presents a "ladder of causation" that distinguishes three levels of causal concepts: associational (level 1), interventional (level 2), and counterfactual (level 3).

Importantly, our results also show that subpopulation-based recourse is the right approach to adopt when assumptions such as additive noise do not hold.