Contextual Morality for Human-Centered Machine Learning