This is a somewhat open-ended question, so let me motivate it a little.
One of the first results in the study of Ordinal Turing Machines includes the following computational characterization of $L$ by Koepke:
Theorem. a set of ordinals is in $L$ if and only if it is computable by an Ordinal Turing Machine with finitely many parameters.
At the same time, $\mathsf{HOD}$ has a "definable power set" characterization: in the construction on $L$, if we replace the first-order definable power set operator with second-order definable power set, then the resulting structure is $\mathsf{HOD}$. This is proved by Myhill and Scott, and is also nicely written up in this MO post.
So a natural question is this: can there be a computational notion $C$ such that something analogous to Koepke's theorem is provable for $\mathsf{HOD}$? By this I have in mind something to the effect of "a set (of ordinals) is in $\mathsf{HOD}$ if and only if it is $C$-computable."
A somewhat trivial candidate would be just oracle-OTM computation. This uses the observation that $\mathsf{HOD}$ is always coextensive with $L[A]$ for some class of ordinals $A$, and we can just relativize Koepke's theorem to $A$. But can we do more?