This post is based upon an interesting property of the Kleisli categories of continuation monads on sufficiently nice base categories, and my blundering about trying to convince myself of why it should be the case.
Stating the property is itself a bit of a mouthful, so we’re going to address that first.
The Property of Interest
Any introductory category theory course will introduce the notion of a cartesian closed category: a category with finite products such that for every object we have an adjunction:
We can also consider the dual notion: a cocartesian coclosed category is a category with finite coproducts such that for every object we have an adjunction:
(I have introduced some non-standard notation for the left adjoint here, as I could not find anything standard).
With that in place, the claim is:
The Kleisli category of a continuation monad on a cartesian closed category with finite coproducts is cocartesian coclosed.
Not very snappy, but the point is that we get a “naturally occurring” source of cocartesian coclosed categories. Of course, we can get lots by taking the opposite of a cartesian closed category, but superficially this looks like they arise more directly.
Aims and Motivation
This observation had been “doing the rounds” and appears as an example on the NLab. My interest arose as follows:
- I did not really get along with the Yoneda style proofs I saw of the property. I normally like this proof style, but for this observation the proofs I encountered left things implicit, and I found making the argument precise cumbersome.
- Even if I felt I could get the proof into a form I felt comfortable with, I could not see why this should be true.
Both of the above points are highly subjective. Mathematical proofs often contain implicit aspects to avoid clutter, and reasons “why something should be true” are very individual. More modestly, this post aims to give a different perspective on what is going on.
An Alternative Argument
Fix a cartesian closed category with finite coproducts, and object
within it. We consider the continuation monad:
This monad arises via the following adjunction:
We can use this adjunction to apply a bit of monad theory we have seen before. The Kleisli comparison functor gives us a full and faithful functor:
with action on objects:
Therefore is equivalent to the full subcategory of
with objects of the form
It is then easy to verify the dual of this category is closed under products and exponentials in , using high school properties of exponentials.
Therefore is cocartesian coclosed, and via the equivalence, so is
.
Conclusion
This post applied the Kleisli comparison functor to rephrase a problem about the Kleisli category of the continuation monad in terms of properties of the (dual of the) base category. This alternative argument made me a more comfortable with the original observation, but disappointed that it seems directly about duality after all.
Acknowledgements: I would like to thank Matteo Capucci for very patiently discussing aspects of the original argument with me.