Stuart_Armstrong comments on Counterfactual resiliency test for non-causal models