In counterfactual cases of discrimination, an agent would have treated someone worse had circumstances been different such that instead of being a member of her actual group, she was a member of some other group. The case for considering such cases to be genuine cases of discrimination is bolstered by the fact that we are inclined to say that cases where an agent would have treated someone better had she been a member of another group are discriminatory. But I argue (...) that the cases are relevantly different: in one kind of case, but not the other, a person is harmed. I then consider and reject a number of objections to this suggestion. (shrink)