Abstract

Category theory offers a simple and unifying understanding of many of the foundations of modern functional programming. It is well known, for example, that data types can be understood via initial algebras, effects can be understood via monads, and dependent types can be understood via fibrations. However, one area in which category theory has thus far been less successful is in providing an elegant and prescriptive understanding of parametricity. Theories such as dinaturality and strong dinaturality have been proposed but they are, unfortunately, insufficient to fully capture all key aspects of parametricity.

I will do my best to rectify this situation by offering an alternative categorical perspective on parametricity. This alternative perspective is based on fibrations. Specifically, I’ll show that the fibrational perspective i) sheds new light on the conceptual essence of parametricity; iI) provides simple and natural formulations of the key constructions of parametricity; and iii) is robust enough not only to cover known models of parametricity, but to suggest new ones as well.