nLab open problems in homotopy type theory

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.

• The 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. One answer is 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 $n^{th}$ universe is a model of HoTT with $(n-1)$ 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 $n\gt 0$, an axiom asserting that there are n-types that are not $(n-1)$-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.

• Define higher inductive types in higher observational type theory.

Homotopy theory and algebraic topology

• Homotopy theory in HoTT?

• Similarly to the torus, consider the projective plane, Klein bottle, … as discussed in the book (sec 6.6). Show that the Klein bottle is not orientable. (This requires defining “orientable”.)

  • This also requires defining what a surface is.

• Calculate more homotopy groups of spheres.

• Show that the homotopy groups of spheres are all finitely generated, and are finite with the same exceptions as classically.

  • The classical proof requires Hurewicz, and now that spectral sequences are around should be possible.

• Define the Toda bracket.

• Prove that $n$-spheres are $\infty$-truncated.

• Prove that $S^2$ is not an $n$-type.

• Define the/a delooping of $S^3$.

• Can we verify computational algebraic topology using HoTT?

• Bott periodicity

• Develop synthetic stable homotopy theory

Higher algebra and higher category theory

• Define semi-simplicial types in single-level non-modal dependent type theory without either type telescopes or internal parametricity, 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.

• Eric Finster, Towards Higher Universal Algebra in Type Theory

• Construct the first 8 $\mathbb{N}$-indexed families of $\infty$-groups in the Whitehead tower of the orthogonal groups.

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 Freyd's adjoint functor theorem?

• Is there a predicative and constructive abstract definition of the category of real Hilbert spaces, in a similar fashion as what Chris Heunen and Andre Kornell did in classical mathematics in their article (Axioms for the category of Hilbert spaces)? Solér’s theorem, which is used in their proof, is only valid in classical mathematics.

In cohesive homotopy type theory

Problems in cohesive homotopy type theory:

• Suppose a discrete cohesive type $A$ comes with a tight apartness relation. Then, assuming crisp excluded middle and axiom C1, is the tight apartness relation on $A$ a crisp relation?

• Suppose a cohesive type $A$ comes with a tight apartness relation. Then, assuming crisp excluded middle and axiom C1, is the tight apartness relation on $\flat A$ in the cohesive mode decidable?

Problems in real-cohesive homotopy type theory:

• Assuming the axiom of real cohesion, is the shape of every subset of the Dedekind real numbers a set?

• Assuming the axiom of real cohesion and crisp excluded middle, prove that the analytic Markov's principle is cohesively true.

• Assuming the axiom of real cohesion and crisp excluded middle, does the full fan theorem hold cohesively? (i.e. is Cantor space a spatial locale?)

• What is the $(\infty,1)$-topos theoretic interpretation of the axiom of real cohesion?

• Prove the approximate intermediate value theorem in Mike Shulman‘s model of real-cohesive homotopy type theory without resorting to crispness.

• Construct a real-cohesive homotopy type theory in multimodal dependent type theory a la GKNB21 and prove the results of Shulman16 in the multimodal dependent type theory

See also the commented list of problems at:

• Some thoughts on the future of modal homotopy type theory (2015):

  • solved problems

  • open problems

In cubical type theory

• Does cubical type theory with regularity have canonicity?

• Does cubical type theory with regularity have an algorithm to compute the canonical forms of closed terms??

• Does cubical type theory with regularity have normalization?

• Does cubical type theory with regularity have an algorithm to compute normal forms?

In weak type theory

• Does univalence imply function extensionality for types in the universe in weak type theory?

• Is (the homotopy type theory based upon) Martin-Löf type theory conservative over (the homotopy type theory based upon) weak Martin-Löf type theory?

• How much of the HoTT book could be done in weak type theory?

• Does weak type theory have homotopy canonicity and normalization?

• What is the categorical semantics of the homotopy type theory based upon weak type theory, with all higher inductive types and weakly Tarski univalent universes?

• Is weak function extensionality equivalent to function extensionality in weak type theory?

• Does product extensionality hold in weak type theory? Namely, given types $A$ and $B$ and elements $a:A \times B$ and $b:A \times B$, is the canonical function $\mathrm{idtoprojectionids}:(a =_{A \times B} b) \to (\pi_1(a) \simeq \pi_1(b)) \times (\pi_2(a) \simeq \pi_2(b))$ an equivalence of types?

• Is function extensionality still provable in weak cubical type theory?

Formalization

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

• Formalize semi-simplicial types in homotopy type theory.

• 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 some language, such as 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.

Other lists of open problems

• Coquand’s five open problems: Coquand listed five open problems here

• Open problems stated at the HoTT2019 Summer School HoTT2019 Summer School open problems list

• Interpretation of Coq in higher toposes.

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.

• There is a model structure on semi-simplicial sets.

• Prove that the torus (with the HIT definition involving a 2-dimensional path) is equivalent to a product of two circles. See Sojakova’s proof: torus.pdf. A shorter formalized proof is here

• Construct the Hopf fibration. (Lumsdaine gave the construction in 2012 and Brunerie proved it was correct in 2013.)

• Construct the “super Hopf fibration” (AKA quaternionic Hopf fibration) $S^3 \to S^7 \to S^4$. (Buchholtz and Rijke solved this early 2016 through a homotopy version of the Cayley-Dickson construction.)

• Prove the Seifert-van Kampen theorem. (Shulman did it in 2013.)

• Construct Eilenberg–MacLane spaces and use them to define cohomology. (Licata and Finster did it in 2013, written up in this paper (pdf).)

• Define the Whitehead product. Guillaume Brunerie did this in 2017, written up in this paper. He also gives a constructive proof of $\pi_4( S^3)= Z/n Z$ for some $n$ and a construction of the reduced James product as a HIT and a homotopy colimit. It is also proven that $\Omega \Sigma X \simeq J(X)$ for some pointed connected type $X$. All of these constructions can be found in detail in his thesis.

• Does having an interval type with only typal beta conversion rules imply function extensionality in Martin-Löf type theory? If yes, does the above still hold if the function types only have a typal eta conversion rule? Answered in the negative in this section about the relation between the interval type and function extensionality: where judgmental beta conversion was used in the original proof becomes function extensionality in the typal case, resulting in circularity.

• Is the propositional truncation of the boolean domain, both of which only have typal computation rules rules, the interval type? The case with judgmental computation rules was done here. Answered in the positive in this section about the relation between the interval type and propositional truncations: one uses the definition of the interval type with a function $j:\mathbb{2} \to \mathbb{I}$ and use double induction on $\mathbb{2}$ to inductively define a function $f:\prod_{a:2} \prod_{b:2} j(a) =_\mathbb{I} j(b)$.

• Define the Hurewicz map and prove the Hurewicz theorem. A proof of the Hurewicz theorem in HoTT can be found here.

• What is the loop space of a wedge of circles indexed by a set without decidable equality? It is a set, and can be shown to be using the zigzag construction.