Learning Global Transparent Models consistent with Local Contrastive Explanations

There is a rich and growing literature on producing local contrastive/counterfactual explanations for black-box models (e.g.