The advent of laboratory experiments in economics over the last few decades has produced an enormous literature devoted to describing, testing and modeling economic and social behavior. Measured by publications and citations, the development of social preference models to capture decisions motivated by fairness and other social criteria, is one of the success stories in this literature. But with this success, and maybe even because of it, controversies have arisen about what the models can and cannot do. In this note, we comment on some of these debates. Our main theme is that descriptive models of behavior should be judged with respect to their usefulness. This is often neglected, partly because there are no accepted measures and tests for the usefulness of a model, while standard procedures to test whether a model is true are readily available. A model that does not capture a grain of truth‘ is unlikely to be useful; however, the relationship is not monotonic in that a truer‘ model is not necessarily a more useful model.