A 2-Categorical Study of Graded and Indexed Monads

TL;DR

This paper develops a 2-categorical mathematical framework for graded and indexed monads, which model parameterized computational effects, providing foundational structures like Eilenberg--Moore and Kleisli objects.

Contribution

It introduces four 2-categories where graded and indexed monads are characterized as monads, and explicitly constructs their Eilenberg--Moore and Kleisli objects, advancing the theory of parameterized effects.

Findings

Established 2-categorical structures for graded and indexed monads.

Constructed Eilenberg--Moore and Kleisli objects for these monads.

Provided a foundation for unified study of parameterized computational effects.

Abstract

In the study of computational effects, it is important to consider the notion of computational effects with parameters. The need of such a notion arises when, for example, statically estimating the range of effects caused by a program, or studying the ways in which effects with local scopes are derived from effects with only the global scope. Extending the classical observation that computational effects can be modeled by monads, these computational effects with parameters are modeled by various mathematical structures including graded monads and indexed monads, which are two different generalizations of ordinary monads. The former has been employed in the semantics of effect systems, whereas the latter in the study of the relationship between the local state monads and the global state monads, each exemplifying the two situations mentioned above. However, despite their importance, the mathematical theory of graded and indexed monads is far less developed than that of ordinary monads. Here we develop the mathematical theory of graded and indexed monads from a 2-categorical viewpoint. We first introduce four 2-categories and observe that in two of them graded monads are in fact monads in the 2-categorical sense, and similarly indexed monads are monads in the 2-categorical sense in the other two. We then construct explicitly the Eilenberg--Moore and the Kleisli objects of graded monads, and the Eilenberg--Moore objects of indexed monads in the sense of Street in appropriate 2-categories among these four. The corresponding results for graded and indexed comonads also follow. We expect that the current work will provide a theoretical foundation to a unified study of computational effects with parameters, or dually (using the comonad variants), of computational resources with parameters, arising for example in Bounded Linear Logic.