Heterogeneity in preferences is important in many domains. However, many existing approaches to studying this phenomenon empirically come with trade-offs for the researcher. For example, parametric approaches often cannot be falsified or distinguish between the predictions of other models. Additionally, while revealed preference methods may identify the existence of a set of types, further economic analysis can be difficult to perform due to the non-parametric nature of these approaches. To complement such methods, this paper introduces a semi-parametric approach to studying preference heterogeneity that is grounded in economic theory, generates falsifiable empirical predictions, and yields economic content beyond a yes/no identification of diverse preferences. I present several empirical applications to demonstrate the advantages of this approach.