Understanding Monads and Beck's Monadicity Theorem

Monads arise in category theory as “monoids in the category of endofunctors,” a viewpoint that underlies both abstract mathematics and functional programming languages such as Haskell [2]. A central result about monads is Beck’s monadicity theorem, which gives precise conditions under which a functor can be recovered from its associated monad [1].

Key takeaways

• A functor is monadic iff it has a left adjoint, reflects isomorphisms, and preserves certain coequalizers [1].
• The theorem has variations that relax the coequalizer requirements, e.g., using reflexive pairs [1].
• In programming, the maybe monad adds a single “exception” value to a type, modeling partial functions [3].
• The maybe monad is not idempotent, so it does not form a modality in the sense of homotopy type theory [3].

Beck’s Monadicity Theorem and Its Variants

Beck’s monadicity theorem characterises monadic functors by three criteria: the existence of a left adjoint, the reflection of isomorphisms, and the preservation of coequalizers of U‑split parallel pairs [1]. When these hold, the functor is equivalent to the forgetful functor from the category of algebras for its associated monad. Several refinements exist. One “crude” version replaces the requirement on U‑split coequalizers with the existence of coequalizers of reflexive pairs, still guaranteeing monadicity [1]. Another formulation demands that every diagram sent by U to a split coequalizer in the target category already be a coequalizer in the source, i.e., U creates those coequalizers [1]. A stricter version distinguishes strictly monadic functors, where the comparison functor is an isomorphism rather than merely an equivalence, and requires uniqueness of the created coequalizers [1].

These categorical conditions have concrete applications. For instance, the forgetful functor from semigroups to sets is monadic, yet it fails to preserve arbitrary coequalizers, illustrating why the theorem’s coequalizer restrictions are essential [1]. Similarly, the powerset functor from Setᵒᵖ to Set is monadic, a fact that extends to any topos and underpins the existence of finite colimits in such categories [1].

The Maybe Monad as an Illustrative Example

In type‑theoretic settings, the maybe monad adds a distinguished “exception” element to any type X, yielding the type Maybe X = X + unit [3]. Its operations include unit-Maybe (injecting a value), exception-Maybe (the extra point), and extend-Maybe (a bind‑like operator that propagates exceptions) [3]. The maybe monad exemplifies a monad that is not a modality because it is not idempotent; applying the monad twice does not collapse to a single application [3]. This distinction matters in homotopy‑type‑theoretic contexts where modalities correspond to idempotent monads.

Why it matters