Prominent ethical and policy issues such as affirmative action and female enrollment in science and engineering revolve around the idea that diversity is good. However, a precise definition is seldom provided. I show that diversity may be construed as a factual description, a craving for symmetry, an intrinsic good, an instrumental good, a symptom, or a side effect. These acceptions differ vastly in their nature and properties. Some are deeply mistaken and some others cannot lead to concrete policies. It is (...) thus necessary to clarify what one means by ‘diversity.’ It may be a neutral description of a given state; but this is insufficient to act. The idea that there should be the same representation in a specific context as in the overall population is both puzzling and arbitrary. Diversity as intrinsic good is a mere opinion, which cannot be concretely applied; moreover, the most commonly invoked forms of diversity are not intrinsically good. On the other hand, diversity as instrumental good can be evaluated empirically and can give rise to policies, but these may be very weak. Finally, symptoms and side effects are not actually about diversity. I consider the example of female enrollment in science and engineering, interpreting the various arguments found in the literature in light of this polysemy. (shrink)