For each $R$ an object of a cartesian closed category, there is a monad $\mathrm{Cont}_R(A) = [[A,R],R]$, the continuation monad.

If $M$ is a strong monad, we can find for each pair of objects $A$ and $B$ a morphism $MA \times [A,MB] \to MB$, by using strength and then evaluating the exponential and then joining the result, so we can find a morphism $MA \to [[A,MB],MB]$, i.e. $MA \to \mathrm{Cont}_{MB}(A)$.

I think this is a monad morphism: I verified that the two ways of getting from $A$ to $[[A,MB],MB]$ were equal, drew the diagram for the commutation with $\mu$ and got a headache. In any case I believe that it works out fine.

So, roughly speaking we have a monad morphism from any (strong) monad to the continuation monad, which prompts me to say "aha! we have a terminal object!". However, there are two problems:

  • The monad morphism is $M \Rightarrow \mathrm{Cont}_{MB}$; crucially, the codomain depends on $M$, so there's not one single terminal object candidate to point to.
  • I wouldn't know where to even start proving that (or if!) this morphism is unique.

So, what's going on here, categorically speaking? Is this phenomenon actually expressible as a suitable universal property?

  • 1
    $\begingroup$ The terminal monad is actually, funnily enough, the constant $1$. $\endgroup$ – Zhen Lin Jul 6 '14 at 18:21
  • $\begingroup$ That's not so surprising, I suppose. But there's still something going on here. $\endgroup$ – Ben Millwood Jul 6 '14 at 22:50
  • 1
    $\begingroup$ I have a hunch that given a monad $T$ and an object $R$, a monad morphism $T \rightarrow \mathrm{Cont}_R$ is the same thing as a $T$-algebra structure on the object $R$. Thoughts/comments? $\endgroup$ – goblin GONE Dec 17 '16 at 23:16
  • $\begingroup$ I think $\operatorname{Cont}_R$ along with $eval : \operatorname{Cont}_R(R) \to R$, $f \mapsto f (\operatorname{id}_R)$, would be terminal in the category of monads $M$ with a distinguished $M$-algebra structure on $R$. Similar to the observation from goblin's comment. $\endgroup$ – Daniel Schepler Feb 22 '18 at 20:02

One observation I have made since asking this question: $\mathrm{Cont}_{MB}$ frequently has non-trivial automorphisms, and so cannot be a genuine terminal object. Any automorphism of $M$ itself, say $\alpha : M \Rightarrow M$, gives rise to an automorphism which I'll call $\overline{\alpha}$ of $\mathrm{Cont}_{MB}$ via \begin{aligned} \overline{\alpha} \ &: \mathrm{Cont}_{MB} \Rightarrow \mathrm{Cont}_{MB} \\ \overline{\alpha}_C \ &: [[C, MB], MB] \to [[C, MB], MB] \\ \overline{\alpha}_C(f) \ &= k \mapsto \alpha_B^{-1}(f(\alpha_B \circ k)) \end{aligned} I admit I haven't gone through the details, but it seems intuitively like it must work.

| cite | improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.