Most
of these papers describe the situation with phrases like "we are
working
in the internal language of a category with finite limits" or
an
elementary topos with NNO, or in CZF, and by an "abuse of language"
we
interpret "for all x there exists a y" as referring to the giving
of
a function assigning a y to each x. But wouldn't it be more
precise
and less abusive to just work in dependent type theory with
Sigma
and Id types, and sometimes Pi and Nat, and use the untruncated
propositions-as-types
logic where "for all x there exists a y"
literally
means Pi(x) Sigma(y) and therefore (by the "type-theoretic
principle
of non-choice") automatically induces a function assigning a
y
to each x? That would also allow asking and answering the question
of how much UIP is required -- do these model structures exist in
HoTT?
Thank you for your email.
Your suggestion of working in a dependent type theory is interesting. I am not sure what kind of dependent type theory would be sufficient to develop these papers and what would be the best approach to the formalization (e.g. via sets-as-hsets or via sets-as-setoids).
Regarding the dependent type theory, apart from basic rules, I guess one would need:
- some extensionality,
- propositional truncations,
- pushouts,
- some inductive types (for the instances of the small object argument)
- at least one universe (cf. quantification over all Kan complexes).
One could then keep track explicitly of which existential quantifies are to be left untruncated and which ones can be truncated, and then see if everything can be done in HoTT.
Is this the kind of thing you had in mind?
Another approach to avoiding the abuse of language, suggested by Andre’ Joyal, is to develop a theory of “split” weak factorisation systems, i.e. weak factorisation systems in which one has a given choice of fillers, and work with them. This would be a
variant of the theory of algebraic weak factorisation systems. We are working on that.
With best wishes,
Nicola
PS The first link in my email was incorrect. Simon Henry’s paper "
Weak model categories in classical and constructive mathematics” is available at https://arxiv.org/abs/1807.02650.