Conformal prediction + causality = winning 🏆 combination. Great to see the original transductive form of conformal prediction in action.
”Conformal Counterfactual Inference under Hidden Confounding” another great paper from ByteDance Research.
“Personalized decision making requires the knowledge of poten- tial outcomes under different treatments, and confidence intervals about the potential outcomes further enrich this decision-making process and improve its reliability in high-stakes scenarios.
Predicting potential outcomes along with its uncertainty in a counterfactual world poses the foundamental challenge in causal in- ference.
Existing methods that construct confidence intervals for counterfactuals either rely on the assumption of strong ignorabil- ity that completely ignores hidden confounders, or need access to un-identifiable lower and upper bounds that characterize the dif- ference between observational and interventional distributions.
In this paper, to overcome these limitations, we first propose a novel approach wTCP-DR based on transductive weighted conformal prediction, which provides confidence intervals for counterfactual outcomes with marginal converage guarantees, even under hidden confounding.”
#conformalprediction
#causality