Homotopy Type Theory
open problems (Rev #39, changes)

Showing changes from revision #38 to #39: Added | Removed | Changed

A list of (believed to be) open problems in homotopy type theory. To add more detail about a problem (such as why it is hard or interesting, or what ideas have been tried), make a link to a new page.

Semantics

  • Construct a model of type theory in an (infinity,1)-topos. What remains open is the issue of weak Tarski universes.

  • Construct an (infinity,1)-topos?, or precisely an elementary (infinity,1)-topos, out of (the syntactic category of) a system of type theory. Kapulkin 1507.02648 has shown that we obtain a locally cartesean closed quasicategory. Urs Schreiber proposed the following question: “Given an (infinity-)functor from (any flavor of) homotopy type theories to infinity-categories as above, characterize its essential image.”

  • Define a notion of elementary (infinity,1)-topos and show that any such admits a model of homotopy type theory.

  • Semantics for strict resizing rules?. Probably refers to the rules introduced in Voevodsky’s Bergen lecture, as opposed to the ones in the book.

  • Construct (higher) inductive-recursive types in univalent models, or show that they don’t exist.

  • Connect univalence and parametricity? as suggested here.

  • Show that when the construction of the cumulative hierarchy? in chapter 10 of the book is interpreted in the simplicial set model?, it reproduces (up to equivalence) the usual cumulative hierarchy of the ambient ZFC. Conclude that if HoTT+AC proves that the cumulative hierarchy satisfies some statement of set theory, then that statement follows from ZFC; see also this comment and surrounding discussion.

  • The model invariance problem: show that the interpretation of type theory is independent of the model category? chosen to present an (infinity,1)-category?.

  • Treat co-inductive types in HoTT. Part of the problem is that general co-inductive types are not fully understood in ordinary type theory. Using extensionality, we obtain M-types from W-types.

Metatheory

  • “Writing all inductive definitions in terms of Sigma-types and W-types is theoretically possible, but extremely tedious … it’s unknown whether any similar reduction for HITs is possible.” (source)

  • Give a computational interpretation? of univalence and HITs. The cubical set model makes progress on this. PDF

  • Is there a cubical type theory? that describes more exactly what happens in the model of type theory in cubical sets?? Partial solutions: the cubical language, this and this

  • Show in HoTT that the n thn^{th} universe is a model of HoTT with (n1)(n-1) universes. Part of this question is: what is an internal model of HoTT.

  • What is the proof theoretic strength of univalent type theory plus HITs? In particular, can they be predicatively justified? One could start by considering simple classes of HITs; e.g. here before continuing to, say, pushouts, the cumulative sets, the Cauchy reals and the surreals.

  • Is univalence consistent with Induction-Recusion? This would allow us to build a non-univalent universe inside a univalent one. Related to this, higher inductive types can be used to define a univalent universe.

  • Consider the type theory with a sequence of universes plus for each n>0n\gt 0, an axiom asserting that there are n-types? that are not (n1)(n-1)-types; see here. Does adding axioms asserting that each universe is univalent increase the logical strength? Of course, univalence gives funext.

  • Prove or disprove the conjecture that every fibrant type in HTS is equivalent to one definable in MLTT. A potential method of disproof would be to solve the model invariance problem positively for MLTT but negatively for fibrant types in HTS.

Homotopy theory and algebraic topology

Higher algebra and higher category theory

  • Define semi-simplicial types? in type theory, or show that this is not possible. Define Segal space? complete Segal space?.

  • Define a weak omega-category in type theory?.

  • Is it possible to have limits be judgmentally the same as opposite colimits, and simultaneously have (co)limits be judgmentally the same as particular Kan extensions. As partially detailed on an issue on the HoTT/HoTT repo, part of the problem is that the functor ᵒᵖ : (D → C)ᵒᵖ → (Dᵒᵖ → Cᵒᵖ) has a judgmental inverse (the composition is judgmentally the identity functor on objects and morphisms), and, assuming univalence, the precategories (D → C)ᵒᵖ and (Dᵒᵖ → Cᵒᵖ) are propositionally but not judgmentally equal.

Other mathematics

  • What axioms of set theory are satisfied by the HIT model of ZFC? constructed in chapter 10 of the book? For instance, does it satisfy collection or REA?

  • Are any of the stronger forms of the axiom of choice? mentioned in the book (ex 7.8) consistent with univalence?

  • Do the higher inductive-inductive real numbers? from the book coincide with the Escardo-Simpson reals, here and here? Solved by Auke Booij using resizing/impredicativity. Can a predicative treatment be obtained from hSets as a predicative topos (by Rijke/Spitters) and the predicative formalization of Cauchy reals by Gilbert.

  • Can multiplication be defined for the higher inductive-inductive surreal numbers? from the book?

  • Can Rezk complete categories remove the non-constructivity from the applications of the Freyd Adjoint Functor theorem? (FAFT)

Formalization

  • Formalize the construction of models of type theory using contextual categories.

  • Formalize semi-simplicial types.

  • Formalize \infty-groupoids, \infty-categories within HoTT.

  • Formalize what remains to be done from chapters 8, 10, and 11 of the book. In particular, develop the Higher Inductive-Inductive real numbers in Coq (the basics of the surreal numbers have been done).

    In general, this file contains a Coq outline of the book. Instructions for how to contribute are here.

Closed problems

How about keeping a running list of solutions like this:?

Maybe only things not in the HoTT book? Else the list of solved problems gets very long!

OK, here’s the rule: if it was stated here (or on the UF-wiki) as an open problem, then it gets recorded here once it’s solved.

Revision on October 6, 2017 at 04:21:16 by Ulrik Buchholtz. See the history of this page for a list of all contributions to it.