## April 26, 2018

### What Is an n-Theory?

#### Posted by Mike Shulman A few weeks ago at the homotopy type theory electronic seminar I spoke about my joint work with Dan Licata and Mitchell Riley on a “generic” framework for type theories. I briefly mentioned how this fits into a hierarchy of “$n$-theories” for $n=0,1,2,3$, but I didn’t have the time in the talk to develop that idea in detail. However, since that short remark has generated some good discussion and questions at the $n$Forum, and also the question of “$n$-theories” has a substantial history here at the $n$-Category Cafe (e.g. this post from 10 years ago), I thought I would expand on it somewhat here.

Let me clarify at the outset, though, that there may be many possible notions of “$n$-theory”. This is just one such notion, which gives a convenient way to describe what I see happening in our general framework for type theories. Notable features of this notion of $n$-theory include:

1. It identifies 2-theories with a certain kind of doctrine, for those who know that word. (If that’s not you, don’t worry: I’ll explain it).
2. It definitively expects the models of an $n$-theory to form an $n$-category, and not just an $n$-groupoid.
3. It helps clarify the somewhat confusing question of “what is a type theory?”, by pointing out that the phrase “type theory” (and also “internal language”) is actually used in two distinct ways: sometimes it refers to a 1-theory and sometimes to a 2-theory.

Let’s start with a simple example of a theory: the theory of monoids. This is a 1-theory, as you might guess from the fact that monoids form a 1-category. Each individual monoid (like $(\mathbb{N},+,0)$, for instance) is called a model of this theory; thus in general a 1-theory has a 1-category of models. More generally, we can define a monoid object in any category with finite products (say); thus a 1-theory has a 1-category of models internal to any sufficiently structured 1-category.

So far so good, but what exactly is a 1-theory in general? As discussed in this post and its comments, there are several “levels of realization” at which we might consider a theory, but for now I want to focus on the two extreme ends: the fully syntactic and the fully semantic. In a fully syntactic presentation, the theory $T_{Mon}$ of monoids looks like this:

• One generating type $M$.
• Two generating terms $(x:M), (y:M) \vdash (m(x,y) : M)$ and $\vdash (e:M)$.
• Three generating equalities $(x:M),(y:M),(z:M) \vdash (m(m(x,y),z) = m(x,m(y,z)) : M)$ and $(x:M) \vdash (m(x,e)=x : M)$ and $(x:M) \vdash (m(e,x)=x : M)$.

The fully semantic presentation of the theory of monoids is its Lawvere theory $L_{Mon}$ or walking model: the category with finite products freely generated by a monoid object $M$. This has the universal property that for any category $C$ with finite products, functors $L_{Mon}\to C$ that preserve finite products are equivalent to monoid objects in $C$.

The connection between the syntactic and semantic presentations is the following:

• The objects of $L_{Mon}$ are the types obtained by starting with the generating type and freely adding finite product types $A\times B$ and the nullary product $1$.
• The morphisms of $L_{Mon}$ are represented by the terms obtained by starting with the generating terms $m(x,y)$ and $e$ and allowing arbitrary substitutions for variables (i.e. composites of morphisms), such as $m(m(x,x),m(y,m(z,y)))$.
• These terms are then quotiented by the congruence generated by the generating equalities of associativity and unitality.

(The intermediate stages of “realization” I referred to are what you get by, e.g., closing up the generating equalities into a congruence but retaining the special status of the generating terms $m(x,y)$ and $e$, and similarly for other places of drawing the line. I want to ignore these not just for expositional clarity, but because when you get to dependent type theory it’s no longer clear to me that there is anywhere “in the middle” to draw the line; see here.)

Now we can say more generally what a 1-theory is (at least, one kind of 1-theory). A syntactic 1-theory has a set of generating types, a set of generating terms (which may involve not only the generating types but those obtained from them by applying finite products), and a set of generating equalities (which may involve not only the generating terms but those obtained from them by substitutions). And a semantic 1-theory is just a category with finite products. Every syntactic 1-theory generates a semantic 1-theory, which is “freely generated by a model”. And every semantic 1-theory can be generated by at least one syntactic 1-theory, such as the “tautological” one whose generating types are all the objects, whose generating terms are all the morphisms, and whose generating axioms are all the equalities. (I’m not being precise about any of this, but it can be made precise.)

[A brief digression]: Note that a given semantic 1-theory can have many different syntactic presentations. For instance, the semantic 1-theory of monoids can be presented either as above in “biased” form with one binary and one nullary operation, or in an “unbiased” form with one $n$-ary operation for all $n$ (and also in many other ways). Nevertheless, the relationship between syntax and semantics is often formulated as an equivalence of categories. Personally, I’m not as fond of such formulations, because in order to make them true one has essentially to define “morphisms of syntactic theories” just so as to make such an equivalence hold, rather than in the “natural” way consisting of functions between the generating types, generating terms, and generating axioms. It’s certainly important that the biased notion of monoid agrees with the unbiased version, but I think it’s also important that the two notions are different in some sense, and to me the natural way to express this is to say that they are two different syntactic theories that generate the same semantic theory. [End of digression]

Now, so far, all of this is only about “finite-product 1-theories”: the syntactic types are generated only by the finite-product constructors $A\times B$ and $1$, the semantic theories are categories with finite products, and the internal models likewise happen in categories with finite products. The notion such as “category with finite products” that we fix in order to give rise to this whole structure is sometimes called its doctrine.

We can then consider other doctrines, such as “monoidal categories”. In this doctrine, the internal models happen in monoidal categories, the semantic theories are monoidal categories, and the syntactic types are generated by tensor product constructors $A\otimes B$ and $I$. In a much more highly structured doctrine like “locally cartesian closed categories” or “toposes”, the semantic theories are such categories, and the syntactic types are generated by higher-order constructors such as dependent products, power-types, universes, and so on. Each choice of doctrine determines a notion of “syntactic 1-theory in that doctrine”, a notion of “semantic 1-theory in that doctrine”, and the above sort of relation between them.

But, how do we define a “doctrine” in general? We have lots of examples, but without a general structure for them we have to develop the general theory for each such doctrine separately. This is clearly an undesirable state of affairs.

Along the way to answering this question, note first that the semantic 1-theories in any doctrine form a 2-category — for instance, the 2-category of categories with finite products, or the 2-category of monoidal categories, or the 2-category of toposes. Thus, it is natural to guess that a doctrine will be a sort of “2-theory” which has a 2-category of “models”, and that these models coincide with semantic 1-theories in that doctrine/2-theory. We might even leap to a general principle such as:

The models of an $(n+1)$-theory are the semantic $n$-theories in that $(n+1)$-theory (regarded as an “$n$-doctrine”, i.e. a language in which to write $n$-theories).

To see whether this makes sense, let’s do a bit of negative thinking. In what sense can we consider the models of a 1-theory as “semantic 0-theories”? In particular, what is a syntactic 0-theory?

Well, note that a syntactic 1-theory can be regarded as a kind of presentation of a semantic 1-theory: we specify the generating objects, generating morphisms, and relations that they have to satisfy. So a syntactic 0-theory should be a suitable kind of presentation of a semantic 0-theory, i.e. of a model.

For instance, in the 1-theory (= 0-doctrine) of monoids, a syntactic 0-theory is a monoid presentation: a set of generating elements together with relations between formal products of generators. Every monoid presentation $(G, R)$ presents a monoid $L_{G,R}$ in the usual way, i.e. every syntactic 0-theory gives rise to a semantic 0-theory. Moreover, this presented monoid has a universal property: monoid homomorphisms $L_{G,R} \to M$, for any monoid $M$, are equivalent to “internal models” of $(G,R)$ in $M$, i.e. interpretations of each generator $x\in G$ as an element of $M$ such that the relations in $R$ hold in $M$.

Note that although we are speaking of internal models of a syntactic 0-theory in an arbitrary semantic 0-theory (in the same 0-doctrine = 1-theory), these semantic 0-theories are not themselves arbitrary internal models of the 1-theory in question, but specifically models in the “canonical” 1-category $Set$. Thus our principle above should really be refined a bit to

The models of an $(n+1)$-theory in the $(n+1)$-category $n Cat$ are the semantic $n$-theories in that $(n+1)$-theory.

How low can we go? Well, the canonical 0-category is $(-1)Cat$, i.e. the (po)set $\Omega$ of truth values. This is, indeed, a monoid under the operation “and” (the decategorification of the cartesian product in $Set$). Thus, given a 0-theory $M$ (a monoid), the semantic $(-1)$-theories in $M$ (regarded as a $(-1)$-doctrine) are the models of $M$ in $(-1)Cat$, i.e. the monoid homomorphisms $M\to \Omega$.

Since $\Omega$ is the subobject classifier (in $Set$), a function $M\to \Omega$ is the same as a subset of $M$. And asking that such a function is a monoid homomorphism says precisely that this subset is a submonoid, i.e. contains the identity element $e\in M$ and is closed under the multiplication. Thus, the semantic $(-1)$ theories in the 0-theory $M$ are the submonoids of $M$; a statement which, if not obviously correct, is at least not obviously incorrect. [Edit: A commenter points out below that this is wrong: there is an extra condition on such submonoids.]

(As usual, the ladder bottoms out here: the “canonical” $(-1)$-category is $(-2)Cat$, namely the truth value “true”, which is the terminal object; thus there are no nontrivial $(-2)$-theories.)

Since negative thinking seems to justify our principle, let’s go back to climbing upwards: what is a 2-theory? We should expect to have both syntactic and semantic 2-theories. A semantic 2-theory will be, in particular, a 2-category, just as a semantic 1-theory (e.g. a Lawvere theory) is, in particular a 1-category. But it will be a structured 2-category, and a syntactic 2-theory will be a way of “presenting” such a structured 2-category in terms of generating objects, morphisms, 2-morphisms, and relations.

What kind of 2-categorical structure will semantic 2-theories have? Well, that depends on the 3-theory in which they live. (-: (-: (-:

The simplest thing to do would be to categorify our running example. In this case, the 3-theory would have as its models (in $2 Cat$) 2-categories with finite products. These are then the semantic 2-theories in this 3-theory (= 2-doctrine). And a syntactic 2-theory in this 3-theory is a presentation of a particular such 2-category.

Below we’ll mention some other possible 3-theories, but first I want to talk some more about syntax. So far, this discussion has been mainly semantic, with comments about the syntax limited to the vague term “presentation”. But we can be more explicit about what the syntax looks like, especially when we consider varying two adjacent levels simultaneously, such as $n=1$ and $n=2$.

A particular 2-theory specifies the “doctrine” in which we write 1-theories, which syntactically means the rules of the deductive system in which we formulate those 1-theories. For instance, the 2-theory of “categories with finite products” specifies rules like

$\frac{A \; type \qquad B \;type}{A\times B \; type}$

saying that we can form cartesian product types, together with their usual introduction/elimination/computation rules, for instance the pairing rule

$\frac{\Gamma\vdash a:A \qquad \Gamma \vdash b:B}{\Gamma\vdash (a,b) :A\times B}$

A particular 1-theory in this doctrine, on the other hand, specifies generating types, terms, and axioms without varying the rules. The theory of monoids, the theory of groups, the theory of rings, the theory of a module over a ring, and so on — all of these can be specified by generating types, terms, and axioms without varying the rules about finite product types.

A different 2-theory specifies a different doctrine, with corresponding different rules. For instance, the 2-theory of “symmetric monoidal categories” specifies rules like

$\frac{A \; type \qquad B \;type}{A\otimes B \; type}$

for tensor product types, together with different introduction/elimination/computation rules, such as a “linear” pairing rule

$\frac{\Gamma\vdash a:A \qquad \Delta \vdash b:B}{\Gamma,\Delta\vdash (a,b) :A\otimes B}$

This corresponds to tensoring together two morphisms with different domains and codomains (the comma in “$\Gamma,\Delta$” is a “judgmental” version of the tensor product $\otimes$), in contrast to how the pairing rule of cartesian products uses the universal property for a pair of morphisms with the same domain.

We can then formulate particular 1-theories in this doctrine, such as the theories of monoids, comonoids, bimonoids, Hopf algebras, etc. — but not the theories of groups or rings, since these require duplication of variables. (Note that the “same” theory of monoids can be formulated in both doctrines. Indeed, the theory of monoids in the doctrine of finite products is “freely extended” from the theory of monoids in the doctrine of monoidal categories. This sort of “change of doctrine” is another interesting direction to think about, but tangential to the point for now.)

Now I think we can see the last point I mentioned before the fold about the ambiguity in the noun phrase “type theory”. Sometimes when people say “theory” they mean something like the theory of monoids, i.e. a 1-theory consisting (syntactically) of generating types, terms, and axioms. For instance, when people say things like “the 2-category of Martin-Löf type theories is equivalent to the 2-category of locally cartesian closed categories”, they mean “type 1-theories”. But other times when people say “type theory” they are referring to a 2-theory, consisting (syntactically) of rules: for instance, the phrase “Martin-Löf type theory” (singular) has this meaning.

Similarly, the phrase “internal language” sometimes refers to the “tautological” syntactic 1-theory associated to a semantic 1-theory, e.g. “the internal language of the category of reflexive graphs” is a particular 1-theory with one generating type for each reflexive graph, etc.. But other times the phrase “internal language” is used to indicate a particular syntax/semantics adjunction, e.g. “extensional Martin-Lof type (2-)theory is an internal language for locally cartesian closed categories”.

So far so good… but unfortunately the world of syntax is not so simple. Even once we’ve fixed a particular doctrine, there can be many different ways to write down rules in which to formulate syntactic theories. In other words, it’s not really quite right to say that an $(n+1)$-theory specifies the rules for a deductive system for $n$-theories; it’s more correct to say that it specifies the intended categorical structure that such rules should represent. (A syntactic $(n+1)$-theory gives a bit more information than a semantic one here, but still not enough.)

My favorite simple example of this is 0-theories in the 1-theory of monoids (so $n=0$). The 1-theory of monoids specifies the intended structure on a set that a “deductive system for monoids” should represent — namely, a monoid structure. What is a “deductive system for monoids”? Roughly speaking, it’s a method of starting from a monoid presentation $(G,R)$ and generating the entire monoid. We have to first produce a collection of “words” from the generating elements $g$, and then quotient them by the specified relations $R$ as well as the monoid axioms. But what is a “word”? There are many possible answers; here are a few:

1. A word is a finite list of generating elements, like $a b c d$.
2. A word is a finite list of generating elements bracketed in binary and nullary groups, such as $(a b)((()c)d)$.
3. A word is a finite list of generating elements bracketed in groups of arbitrary arity, such as $(a b c)()(d)$.

Answer #1 is, of course, the simplest, and the one we normally use. But answer #3 is all we can produce “naturally” if we start from the semantic 1-theory of monoids, which has exactly one operation of each arity. If we start instead from the “biased” syntactic 1-theory of monoids, which has one binary and one nullary generating term, we can obtain answer #2. Answer #1 involves some deeper analysis of the notion of monoid, since it already incorporates the associativity axiom rather than imposing it afterwards as part of the quotienting. (The proof that answer #1 actually works is, essentially, a form of “cut-elimination”.) Note also that the choice of “notion of word” actually already has to be made before we specify the entire presentation $(G,R)$, since each relation in $R$ has to relate a pair of words.

Similarly up a level, once we fix a 2-theory, there are many different deductive systems that can present 1-theories in that doctrine. For instance, cartesian closed categories can be presented using lambda-calculus, but also combinatory logic. Monoidal categories can be presented as I described above using “intuitionistic linear type theory”, but also a “combinatory” version as in the rosetta stone paper, and even string diagram calculus can be regarded as a way to present monoidal categories.

Any such deductive system has advantages and disadvantages. However, for the purposes of computation — and also, arguably, for the purposes of informal usage as a foundation for mathematics, although that’s a discussion I don’t have space for right now — it’s important to look for a particular collection of advantages, notably the “admissibility” of operations such as cut and substitution, and computational properties such as normalization and canonicity. These sorts of metatheorems are what type theorists spend a lot of their time proving — about each doctrine separately.

Now I can finally describe the programme that Dan, Mitchell, and I are undertaking. Given a particular 3-theory, we aim to describe a particular deductive system for syntactic 2-theories in that 3-theory, and associate to any such syntactic 2-theory $D$ a specific deductive system for syntactic 1-theories in $D$, such that the latter (1) indeed has the correct semantics, i.e. it presents the correct semantic 1-theories, and (2) has the good meta-theoretic properties that type theorists want, such as cut-elimination, normalization, and canonicity. For a particular syntactic 1-theory, the deductive system that we get is not usually identical to any existing deductive system for that 1-theory, but it is almost always related to it by a fairly direct translation (in general, rules of the existing system tend to translate into the successive applications of two or more “more basic” rules in our system).

To the extent that it succeeds, this has the potential to simplify our lives considerably: instead of treating each 2-theory separately, we can obtain them all automatically from a general theorem. Or, rather, from a small collection of general theorems, because (for the present, and also the foreseeable immediate future) we are treating each 3-theory separately.

• The first paper by Dan and myself dealt with the 3-theory of (totally unstructured) 2-categories. Syntactically, this corresponds to “unary type (2-)theories” with judgments such as $x:A \vdash b:B$ with exactly one type to both the left and the right of the turnstile.
• The second paper, with Dan and Mitchell, dealt with the 3-theory of cartesian monoidal 2-categories (actually, for technical reasons we used cartesian 2-multicategories). Syntactically, this corresponds to “simple type (2-)theories” with judgments such as $x:A, y:B, z:C \vdash d:D$ with a finite list of types to the left but exactly one type to the right, and no dependent types.
• At the end of my HoTTEST talk, I reported on our current progress towards a 3-theory that corresponds syntactically to “dependent type (2-)theories”. The semantic version of this seems to be some kind of “comprehension 2-category”.

These three 3-theories are “cumulative”, i.e. any 2-theory in one of the first two can also be formulated in the more expressive ones below it, in the same way that the theory of monoids in the doctrine of monoidal categories can also be formulated in the more expressive doctrine of categories with finite products. So once our current work on dependent type 2-theories is complete, it will (to a certain extent) subsume the previous two papers. However, there are also other 3-theories that remain to be studied. Some, such as the 3-theory of first-order logic, live in between the 3-theories of simple type 2-theory and dependent type 2-theory. But another notable one is:

• There should be a 3-theory that corresponds syntactically to “classical simple type (2-)theories” with judgments such as $A,B,C \vdash D,E$ that allow finite lists of types on both sides of the turnstile. For instance, the 2-theory of $\ast$-autonomous categories (classical linear logic), and also of Boolean algebras (classical nonlinear logic), live in this 3-theory.

We have some ideas about this latter 3-theory, but it’s still rather up in the air. It should subsume the 3-theory of simple type 2-theories, but is incomparable with the 3-theory of dependent type 2-theories (neither subsumes the other), and it’s unclear to me whether there is any 3-theory that subsumes both of them.

Since this is the $n$-Category Café, someone is going to ask the question “what about 4-theories?” or even “what about $\infty$-theories?”, so let me go ahead and answer it preemptively.

In principle, the sort of “interaction between adjacent levels” that we’re interested in — using a 2-theories in a particular 3-theory to describe a deductive system for syntactic 1-theories in such 2-theories — could be generalized to any adjacent levels. For instance, we could imagine using 7-theories in a particular 8-theory to describe a deductive system for syntactic 6-theories in such 7-theories. However, in practice, when dimensions get beyond 2 or 3, human beings seem to do better with some sort of “encoding” of the higher-dimensional structure in terms of lower-dimensional structure.

Homotopy type theory is a good example: despite being ostensibly a theory of $(\infty,1)$-categories, as a type theory it is actually just a 2-theory, or a deductive system for a certain class of 1-theories. Syntactically, this is roughly because it doesn’t include any basic 2-cell judgments; its 2-cells and higher cells are all encoded implicitly using 1-morphisms into path objects. Semantically, this corresponds to the fact that its most “direct” notion of model is actually a kind of structured 1-category (a comprehension category, category with families, etc.); a coherence theorem then (hopefully) relates these to the intended $(\infty,1)$-categorical models.

So, while in principle one could go on to talk about 4-theories, $n$-theories, and $\infty$-theories, it’s not clear to me that anyone will.

To conclude, let me say a bit about how this project relates to modal logic. In a syntactic 2-theory with multiple generating types, the objects of the resulting semantic 2-category are not single structured categories, but diagrams of several categories with functors and natural transformations between them. Thus, the corresponding syntactic 1-theories have several “classes” of types, one for each category. These classes of types are generally called “modes”, type theory or logic with multiple modes is called “modal”, and the functors between these categories are called “modalities”. Thus, modal logics are particular 2-theories, to which our framework applies.

Traditional modal logic involves “logical” modalities such as $\Box$ (“it is necessary that…”) and $\lozenge$ (“it is possible that…”). More recently, in cohesive type theory (first here) we have introduced modalities denoted $ʃ,\flat,\sharp$ that correspond to the monads and comonads on a cohesive topos. Other applications have also appeared or are in the works, such as this paper on internal universes, a type theory for parametrized spectra involving a self-adjoint modality $\natural$, and a type theory for “differential cohesion” involving six modalities $ʃ,\flat,\sharp,\Re,\Im,\&$. I fully expect that in the future such modal type theories will proliferate, so having a general framework to produce them will be very useful.

Posted at April 26, 2018 9:50 PM UTC

TrackBack URL for this Entry:   https://golem.ph.utexas.edu/cgi-bin/MT-3.0/dxy-tb.fcgi/3034

### Re: What is an n-theory?

This is great to see such work coming to fruition, Mike. To think it was around 14 years ago that John Baez was telling me about his idea of modal logic as a form of categorified logic. The only glimpse I can see of some ideas we were knocking about back then is in this comment. The idea of a dependency through two indexing layers was there, as is the idea that there is something modal about this.

Anyway, I’m really looking forward to seeing modal type theory get properly worked out. There should be many uses for philosophy.

Posted by: David Corfield on May 2, 2018 3:57 PM | Permalink | Reply to this

### Re: What is an n-theory?

I agree, it’s exciting to see vague ideas from long ago start to find precise expression! It would be great to work out how some of the discussion back then corresponds to the picture I presented. But I had a bit of trouble following that old discussion, perhaps partly because it referred to older ones (perhaps some offline) that I didn’t have at hand. Where would be a good place to start?

Posted by: Mike Shulman on May 2, 2018 4:59 PM | Permalink | Reply to this

### Re: What is an n-theory?

Well perhaps we didn’t get too far. There’s a snatch of discussion from the ‘2-toposes’ thread here. Then further in the ‘Worrying about 2-logic’ thread, there is John here.

Oh, I see a very neglected nLab page Philosophy has this:

There ought to be a categorified logic, or 2-logic. There are some suggestions that existing work on modal logic is relevant. Blog discussion: I, II, III, IV, V. Mike Shulman’s project: 2-categorical logic.

I think that idea of two layers of dependency was only really discussed verbally.

No doubt Jim Dolan was heavily involved in the initial ideas. I wonder what there is in their ‘doctrines’ work: Doctrines of algebraic geometry and Doctrines.

Posted by: David Corfield on May 3, 2018 10:01 AM | Permalink | Reply to this

### Re: What is an n-theory?

Thanks for all those links! Here’s a “number 0” that’s good to start with. I’d forgotten that I was a part of some of those conversations.

(As an aside, I noticed this comment and ensuing discussion, which in hindsight foreshadows the notion of h-level and the $\Sigma/\exists$ distinction in HoTT.)

A lot of that earlier discussion seems to focus on groupoids rather than categories. Perhaps one reason is the fact mentioned by Todd at the bottom of this post that for a first-order theory $T$ in classical logic with no function symbols, the category $T Mod$ is automatically a groupoid. But generalizing to other doctrines, we shouldn’t expect this to remain true.

Posted by: Mike Shulman on May 3, 2018 7:48 PM | Permalink | Reply to this

### Re: What is an n-theory?

One more glimpse of this early work is at the end of my Categorification as a heuristic device. It’s a shame I have nothing better than a garbled Word document. The book it appeared in was certainly not a bestseller.

Posted by: David Corfield on May 28, 2018 10:14 AM | Permalink | Reply to this

### Re: What is an n-theory?

Posted by: David Corfield on June 10, 2018 10:33 AM | Permalink | Reply to this

### Re: What is an n-theory?

Can you give us a sense of the form of these ‘serious restrictions’?

Basically our approach is that all type-forming operations are required to be given by adjunctions (of some generalized/dependent/multivariable sort). You can omit some of the adjoints if you want (e.g. to describe monoidal categories that aren’t closed, or modal operators that don’t have adjoints), but the adjoints are always potentially there within the formalism.

Moreover, all the adjunctions are, as Neel says, internalizations of judgmental structure, which in categorical terms means that they have universal properties. For instance, the tensor product internalizes the judgmental comma:

$\frac{ A,B \vdash C} {A\otimes B\vdash C}$

which categorically means that instead of monoidal categories we talk about representable multicategories. Similarly, a modal operator like $\Box$ or $\flat$ internalizes a category of “valid/crisp judgment”:

$\frac{ x::A \vdash C} { x:\flat A \vdash C}$

which categorically means instead of pseudofunctors we are talking about (bi)fibrations. In fact, you can view our semantic construction as a somewhat idiosyncratic approach to “generalized multicategories” and their representability.

Posted by: Mike Shulman on June 11, 2018 10:46 AM | Permalink | Reply to this

### Re: What is an n-theory?

I just came across this again.

a somewhat idiosyncratic approach to “generalized multicategories” and their representability.

How broadly are you taking generalized multicategory here?

I see in your Categorical logic from a categorical point of view that you’re looking at a certain kind in §2.6, derived from faithful cartesian clubs, which you develop as type theories in §2.7. But can one develop a type theory more broadly?

Posted by: David Corfield on June 23, 2020 10:25 AM | Permalink | Reply to this

### Re: What is an n-theory?

LSR is significantly more general than faithful cartesian clubs, but less general than an arbitrary monad. Basically it’s anything that can be described by a bifibration over some cartesian multicategory. (FWIW, the type theories obtained directly from LSR are a bit different from the ones I described explicitly for faithful cartesian clubs — they decompose the structural rules into structurality-over-structurality and a 2-cell action — but there’s a translation.)

Posted by: Mike Shulman on June 24, 2020 6:01 AM | Permalink | Reply to this

### Re: What is an n-theory?

Thanks! Would it be possible to characterise bifibrations over cartesian multicategories in terms of a kind of monad?

Posted by: David Corfield on June 24, 2020 12:16 PM | Permalink | Reply to this

### Re: What is an n-theory?

Hmm, I think maybe I worked this out at one point, but I don’t remember the details. Let’s see, just as an operad in $V$ induces a monad on $V$ whose algebras are the $V$-algebras, a cartesian 2-multicategory $M$ should induce a monad on $Cat^{ob(M)}$ whose algebras are the functors $M\to Cat$. Maybe that’s the monad? I’ll probably sit down and work out the details while I’m expanding the categorical logic notes later this year.

By the way, I misspoke: the multicategories correspond not to the bifibrations, but to the local discrete fibrations over $M$. The bifibration conditions are representability/closedness of the multicategory.

Posted by: Mike Shulman on June 24, 2020 10:53 PM | Permalink | Reply to this

### Re: What is an n-theory?

Mike wrote

Basically our approach is that all type-forming operations are required to be given by adjunctions (of some generalized/dependent/multivariable sort).

I was reminded of this comment while having an interesting conversation with Christopher Dorn the other day. He works on manifold diagrams and way they might be used in a geometric approach to $n$-categories.

Turning to logic, I was wondering then whether we might see some sign of the singularities structures appear in a higher form of logic. E.g., where the swallowtail singularity corresponds to coherent biadjunctions, might there be a place for the latter in a higher logic, say in the specification of a higher theory?

Posted by: David Corfield on November 19, 2022 8:59 AM | Permalink | Reply to this

### Re: What is an n-theory?

Perhaps! So far, I think everyone has tried to stick to 1-adjunctions because they’re more convenient and simpler, applying coherence theorems to strictify higher ones. But recently I’ve started to wonder whether modal homotopy type theories might eventually need the mode-theory 2-category to be generalized to an $(\infty,2)$-category, with corresponding rules in the syntax for the higher cells.

Posted by: Mike Shulman on March 15, 2023 2:37 AM | Permalink | Reply to this

### Re: What is an n-theory?

Interesting!

With the announcement of a univalent directed type theory, perhaps we’re seeing something like the internal logic of $(\infty, 2)$-toposes emerge. I was thinking what a modal form of this might look like.

E.g., given those ideas we kicked around about necessity and possibility modalities as (co)monads generated between slices of a $(\infty, 1)$-topos, what is the parallel for slices of an $(\infty, 2)$-topos?

I guess we do see interest in variation over $(0, 1)$-categories, i.e., posets.

Posted by: David Corfield on March 15, 2023 5:26 PM | Permalink | Reply to this

### Re: What is an n-theory?

People have been working on directed type theories for a long time. But yes, whatever approach we settle on, eventually it ought to give an internal logic of $(\infty,2)$-toposes (or $(\infty,n)$ or even $(\infty,\infty)$). One of the hard things about it is that such a category can be expected to have more than one kind of “slice”, depending on whether you take fibrations, opfibrations, or what.

I wasn’t thinking about this in my last comment, though – just about using a small $(\infty,2)$-category to index a modal type theory for a diagram of $(\infty,1)$-toposes.

Posted by: Mike Shulman on March 15, 2023 6:16 PM | Permalink | Reply to this

### Re: What is an n-theory?

Given your slide 20, could we also think in terms of (categorified) hyperdoctrines?

I guess given a 2-topos with nice behaviour of its fibrational slices, we might have something like that.

Posted by: David Corfield on May 2, 2018 4:50 PM | Permalink | Reply to this

### Re: What is an n-theory?

Roughly speaking, I would say that hyperdoctrines (and indexed monoidal categories) are the kind of 1-categorical structures that correspond to 2-theories in the 3-theory of “first-order logic”, which I mentioned briefly as living between the 3-theory of simple type (2-)theories and the 3-theory of dependent type (2-)theories. By the Baez-Dolan microcosm principle (which apparently I forgot to mention in the post!), semantic 2-theories in this 3-theory should be a sort of “2-hyperdoctrine”, by analogy to “cartesian monoidal 2-categories” and “comprehension 2-categories”. Syntactically, this 3-theory has one layer of dependency: there is one layer of types which can’t depend on anything, and then there is a second layer of types (“propositions” in the logical reading) which can depend on types in the first layer, but not on each other.

Posted by: Mike Shulman on May 2, 2018 4:57 PM | Permalink | Reply to this

### Re: What is an n-theory?

Cool post! As an outsider to HTT, it took me a long time to realize one of the points that you bring up: HTT is $2$ theory which is meant to axiomatize the structure of a particular type of category (roughly the structure that one sees in the subcategory of fibrant objects + fibrations in a model category that “presents an $\infty$-topos”). In particular, it is not an $\infty$-theory that axiomatizes the structure of an $\infty$-topos– which is what the slogan “HTT is the internal language of $\infty$ topoi” suggested to me at first.

It seems that one big tensions in higher category theory is that those of us who are algebraically or computationally oriented prefer syntactic theories like HTT which have relatively simple presentations. And now, even though we have the basic tools to talk about $n$-theories, e.g. as $n$ Segal spaces with finite products, these tools are only semantic.

I guess that the reason for this is that the only presentation that exists uniformly is the “tautological one” you mentioned, and there is no reason for a mathematically interesting theory to have a simple presentation: For instance finding a simple syntactic presentation the $(\infty,1)$ theory of $E_2$ algebras is morally equivalent to finding simple $CW$ decompositions for the configuration spaces $\text{syntax error at token R}$ which are compatible with composition– this takes hard work!

Posted by: Phil Tosteson on May 2, 2018 4:58 PM | Permalink | Reply to this

### Re: What is an n-theory?

Yes, I think I agree. It’s not impossible to write down true $\textrm{ Conf}_n \mathbb R\infty$-theories, though. For instance, opetopic type theory is arguably a true syntactic $(\infty,\infty)$-theory.

Cubical type theory is an interesting case. I’m inclined to regard it as still a 2-theory analogous to HoTT, although it does incorporate higher cells somewhat “judgmentally” by using interval variables. This sort of classification may not have a unique answer.

Posted by: Mike Shulman on May 2, 2018 5:07 PM | Permalink | Reply to this

### Re: What is an n-theory?

– and finding those decompostitions is hard!

Unrelatedly, is there currently a $3$ theory $T$ which does for $\infty$ categories what HTT does for $\infty$ groupoids? In other words, models of $T$ are structured $2$ categories, one model should be the $2$ category of quasi-categories, but which also includes other “$\infty$ cosmoi.”Is this the sort of theory that your work with Emily Riehl is developing?

Posted by: Phil Tosteson on May 2, 2018 4:59 PM | Permalink | Reply to this

### Re: What is an n-theory?

No, such a theory doesn’t yet exist. My work with Emily is actually a 2-theory whose models include the 1-category of Reedy fibrant bisimplicial sets, inside which one can describe the Segal types and Rezk types which form a model for $(\infty,1)$-categories.

Posted by: Mike Shulman on May 3, 2018 6:22 PM | Permalink | Reply to this

### Re: What is an n-theory?

I was just wondering whether you have reflected upon taking say the 1-theory of monoids that you described, and replacing judgemental equality by intensional equality? Does this change the level of the theory?

Posted by: Richard Williamson on May 3, 2018 10:22 AM | Permalink | Reply to this

### Re: What is an n-theory?

The 1-theory of monoids that I described lives in the 2-theory of cartesian monoidal categories, wherein there is no notion of “intensional equality”. You can also formulate a 1-theory of monoids inside the 2-theory of intensional type theory; the result is still a 1-theory although it lives in a different 2-theory.

Posted by: Mike Shulman on May 3, 2018 6:15 PM | Permalink | Reply to this

### Re: What is an n-theory?

Yes, I meant that one replaces $(x:M) \vdash (m(x,e)=x:M)$ by $(x:M) \vdash w_{x,e} : Id_{M}(m(x,e), x)$, and so on, in some 2-theory (?) in which one has identity types.

What prompted my question was that you wrote

A syntactic 1-theory has a set of generating types, a set of generating terms (which may involve not only the generating types but those obtained from them by applying finite products), and a set of generating equalities (which may involve not only the generating terms but those obtained from them by substitutions).

and wrote that in a 1-theory one formulates axioms

without varying the rules

whereas the intensional re-formulation of the axioms of a monoid are, to me, rules rather than axioms.

Perhaps you simply meant to restrict your statement to rules involving type constructors, introduction, elimination, and computation?

Posted by: Richard Williamson on May 3, 2018 6:49 PM | Permalink | Reply to this

### Re: What is an n-theory?

I don’t know what you mean; your $(x:M) \vdash w_{x,e} : Id_M(m(x,e),x)$ is a generating term, not a rule.

Posted by: Mike Shulman on May 3, 2018 7:01 PM | Permalink | Reply to this

### Re: What is an n-theory?

What do you consider to be a rule? What if we write it as follows?

$\frac{\vdash x \colon M}{\vdash w_{x,e} : Id_{M}(m(x,e), x)}$

What I mean by a rule is something that given a term in context, prescribes that we can form another term in some other context, and I distinguish this from an axiom by the fact that we are not asserting (judgemental) equality of anything.

Posted by: Richard Williamson on May 3, 2018 7:36 PM | Permalink | Reply to this

### Re: What is an n-theory?

What I mean by a rule is something that given some judgments (the premises) prescribes that we can form some other judgment (the conclusion). Terms-in-context are particular judgments, as are judgmental-equality judgments.

One way to say what I mean by a “generating thingy” or “axiom” is a rule without premises. So generators/axioms can be regarded as rules (and, as such, can be frobnicated into other equivalent (modulo admissibility of substitution) forms that do have premises, such as yours), but not every rule can be regarded as a generator/axiom. From this perspective, “formulating axioms without varying the rules” means “adding only rules without premises”.

It’s helpful to distinguish axioms from (more general) rules; in particular, it’s the only way to distinguish the theory from the doctrine. However, the fact that axioms can be regarded as a degenerate case of rules is also important, and means you can always “build a theory into a doctrine”. In other words, instead of considering the 1-theory $T^1_Mon$ of monoids inside the 2-theory $T^2_FP$ of categories with finite products, you could consider the 2-theory “$Lift(T^1_Mon)$” of “categories with finite products containing a monoid object”. That is, the models of $Lift(T^1_Mon)$ are the models of $T^2_FP$ containing a model of $T^1_Mon$.

There is another interesting general fact here, namely that the monoid of global sections of the generic $T^1_Mon$-model internal to the initial $Lift(T^1_Mon)$-algebra actually coincides with the initial $T^1_Mon$-model in $Set$. Which means that if you had a nice enough syntactic presentation for 2-theories, it could also function as your syntactic presentation of 1-theories by simply incorporating them into the 2-theory in this way. Dan, Mitchell, and I have considered going this route, but so far it’s been difficult to formulate a presentation of 2-theories for which this would approach the niceness of our presentations of 1-theories given a fixed 2-theory (and there are other issues too).

Posted by: Mike Shulman on May 3, 2018 8:02 PM | Permalink | Reply to this

### Re: What is an n-theory?

Nice, thanks for the elaboration!

Which means that if you had a nice enough syntactic presentation for 2-theories, it could also function as your syntactic presentation of 1-theories by simply incorporating them into the 2-theory in this way.

That makes sense, and is a nice observation I think. By coincidence I have been contemplating, on the semantic side, a somewhat similar situation recently.

Posted by: Richard Williamson on May 4, 2018 12:49 AM | Permalink | Reply to this

### Re: What is an n-theory?

That part about semantic $(-1)$-theories of the $0$-theory of a monoid being submonoids put me in mind of when you have a theory in propositional logic given by some atomic sentences and axioms. This gives rise to a Boolean algebra, where models of the theory are given by Boolean algebra morphisms to $\mathbf{2}$.

So say we have $P$, $Q$, $R$, and one axiom $P \wedge Q \to R$, then models pick out certain subsets of the set of atomic states satisfying the axiom.

So what kind of $n$-theory is propositional logic? Since models are Boolean algebras, which form a category, does this make it a syntactic 1-theory? Then a specific presentation of a propositional theory is a 0-theory whose models form the set of semantic (-1) theories.

But isn’t that surprising that first-order logic is a 3-theory, two levels higher?

Posted by: David Corfield on May 3, 2018 12:08 PM | Permalink | Reply to this

### Re: What is an n-theory?

There is indeed a 1-theory whose 1-category of models is the category of Boolean algebras, and such that a 0-theory in that 1-theory is a presentation of a specific Boolean algebra.

However, a single Boolean algebra is also a category in its own right, and so the collection of Boolean algebras is a 2-category (which happens to be a (1,2)-category). Thus there is also a 2-theory whose 2-category of models is the 2-category of Boolean algebras, and such that a 1-theory in that 2-theory is a presentation of a specific Boolean algebra.

The latter presentation corresponds better to what we think of as propositional logic. The 2-theory of Boolean algebras naturally lives in the 3-theory of “classical 2-theories” (mentioned briefly in the post as not yet having been studied), and as such can be presented with good type-theoretic properties such as cut-elimination.

Posted by: Mike Shulman on May 3, 2018 6:10 PM | Permalink | Reply to this

### Re: What is an n-theory?

Also, I meant by “first-order logic is a 3-theory” is that there’s a 3-theory for 2-theories with the “dependency shape” of first-order logic. A specific such 2-theory, like the ordinary first-order classical logic that one learns in a first course on mathematical logic, is itself (like propositional logic) a 2-theory. Its models are some kind of hyperdoctrine. Other 2-theories in that same 3-theory are first-order intuitionistic logic, first-order linear logic, the type theory of indexed monoidal categories, first-order modal logic, etc.

Posted by: Mike Shulman on May 3, 2018 6:12 PM | Permalink | Reply to this

### Re: What is an n-theory?

As you say “change of doctrine” is tangential to this post, but I wonder if we could have a brief glimpse of how things may go.

I imagine that when mapping from one $n$-theory to another, you do so ‘within’ the $n+1$-theory to which they both belong. So we might inject $Th^1_{Mon}$ into $Th^1_{Grp}$ within $T^2_{FP}$.

On the syntactic side, presumably we should heed your warning about “morphisms of syntactic theories” as to how we should rely on

the “natural” way consisting of functions between the generating types, generating terms, and generating axioms.

Perhaps this gives rise to some Morita-equivalence issues.

You’ve already mentioned a map of 2-theories

the theory of monoids in the doctrine of finite products is “freely extended” from the theory of monoids in the doctrine of monoidal categories

So what happens generally with a syntactic morphism (within a 3-theory) between syntactic 2-theories? Such a morphism should act on all the 1-theories written in the domain 2-theory, right?

At some point we should make contact with the concept of a doctrinal adjunction which is taking ‘doctrine’ in the sense of 2-monad. (But note the warning that 2-monads aren’t the correct way to go for doctrines.) So that’s about lifting adjunctions of semantic 1-theories from one 2-theory to another?

Finally, we would have syntactic morphisms between 3-theories. You’ve already given us some idea of relations between five of these: unary, simple, first-order, dependent, classical simple. (Is that the best name for the latter?) So morphisms here bring about transformations of lower level theories.

Posted by: David Corfield on May 4, 2018 8:49 AM | Permalink | Reply to this

### Re: What is an n-theory?

Comments are a great place for discussing tangential things! I haven’t thought about change-of-doctrine very deeply in this context, but I can shoot from the hip.

I imagine that when mapping from one $n$-theory to another, you do so ‘within’ the $n+1$-theory to which they both belong.

Yes, that’s definitely the most direct version. I could imagine some kind of “dependent mapping” that first extends an $n$-theory along a map of $(n+1)$-theories and then maps it into an $n$-theory in a different $(n+1)$-theory, but that decomposes into two same-theory mappings at different levels.

So what happens generally with a syntactic morphism (within a 3-theory) between syntactic 2-theories? Such a morphism should act on all the 1-theories written in the domain 2-theory, right?

Yes, I think so. For instance, the morphism from the 2-theory of monoidal categories to the 2-theory of cartesian monoidal categories is of the simplest sort that just “adds structure”, namely diagonals and projections, which syntactically corresponds to the structural rules of contraction and weakening. So any syntactic 1-theory in the doctrine of monoidal categories can almost without modification be interpreted in the doctrine of cartesian monoidal categories, simply be neglecting to make any use of these rules.

In general, a theory morphism doesn’t have to be injective. For instance, there is a 2-theory morphism from cohesive type 2-theory to ordinary type 2-theory that collapses the separate crisp and cohesive contexts into the one ordinary context. Thus, any 1-theory in cohesive type 2-theory can be interpreted as a 1-theory in ordinary type 2-theory by simply forgetting all the crispness annotations.

It’s an interesting fact (which has, of course, been noted by many people) that on the side of syntax, the “natural” action of doctrine morphisms is “covariant” in this sense, while on the side of semantics, the “natural” action is contravariant: any cartesian monoidal category is in particular a monoidal category, and any topos can be equipped with the identity cohesive structure over itself. In the semantic world, the covariant action is more complicated: you have to “build the free cartesian monoidal category on a monoidal one” by freely adding diagonals and projections. Similarly, in the syntactic world, the contravariant action is more complicated: you have to express a theory that uses contraction and weakening in a language that lacks them by “putting them in by hand” as generators. These two actions are presumably related in some way “over” the adjunction between syntax and semantics, which ought to be obvious but is too much for my brain on a Friday afternoon at the end of the semester.

Finally, we would have syntactic morphisms between 3-theories.

Right. When I talked about “cumulativity” of 3-theories “subsuming each other” in the post, I must formally have meant to refer to morphisms of 3-theories, allowing us to naturally extend 2-theories “up the ladder”.

Posted by: Mike Shulman on May 4, 2018 11:43 PM | Permalink | Reply to this

### Re: What is an n-theory?

We could do with a clever notational device to specify trails of $n$-theories. There you are giving a monoid presentation and we might want to know which 1-, 2- and 3-theory you’re working in.

What of the conversation we were having with Urs about what happens when we take a subcategory of a semantic $n$-theory. As you pointed out, we may not be able to provide a corresponding syntactic $n$-theory, at least not in the same $n+1$-theory.

I wonder if there will be reasonably general things to say about when such a move is possible in the same $n+1$-theory? Or when a shift depending on some morphism of $n+1$ theories is possible. There could also be two step decompositions, as you mention.

Posted by: David Corfield on May 5, 2018 10:07 AM | Permalink | Reply to this

### Re: What is an n-theory?

Well, one way of specifying a subcategory of a given semantic $n$-theory is by adding axioms to its corresponding syntactic $n$-theory. The objects of the subcategory will then be the models that satisfy the additional axioms.

Posted by: Mike Shulman on May 8, 2018 11:34 PM | Permalink | Reply to this

### Re: What is an n-theory?

Could we add higher-order logic as a sixth 3-theory?

Then where would that sit? I see Bart Jacobs speaks of Translating dependent type theory into higher order logic. Perhaps this concerns specific 2-theories in their respective 3-theories, or is it possible to see the translation as resulting from a morphism of 3-theories?

Posted by: David Corfield on May 23, 2018 9:34 AM | Permalink | Reply to this

### Re: What is an n-theory?

Higher-order logic is a funny beast from a doctrinal perspective. I’m not positive, but I think it can actually be expressed as (a class of) 2-theories within the 3-theory that I’m calling first order logic: just add rules specifying an a-la-Tarski universe $\Omega$ of propositions.

I think of a 3-theory as specifying the arity and dependency structure. The 3-theory of “first order logic” has a base whose arity is many-to-one with no internal dependency and then another layer that depends on that. (The upper layer could be many-to-one or many-to-many, so actually there are two different 3-theories for first-order logic, just as there are two different 3-theories for propositional logic / simple type theory. The many-to-one and many-to-many versions are traditionally called “intuitionistic” and “classical”, though that’s confusing too.) Higher-order logic has this same arity and dependency structure; it just happens to have a universe as well. (So maybe “first order logic” is a bad name for this 3-theory…)

Posted by: Mike Shulman on May 24, 2018 2:28 PM | Permalink | Reply to this

### Re: What is an n-theory?

Could we see HOL as a lifted 2-theory of a 1-theory, $X$, in the 2-theory of first order logic, $T^2_{FOL}$, so that $T^2_{HOL} =Lift(T^1_{X})$?

Something perhaps that makes sense semantically of triposes as first-order hyperdoctrines equipped with something extra?

I’d like to hear more of this lifting idea. Presumably it’s not restricted to act between the 1- and 2-levels.

Posted by: David Corfield on September 26, 2020 5:02 PM | Permalink | Reply to this

### Re: What is an n-theory?

Well, I don’t think the fact that some type is a subobject classifier can be expressed using first-order axioms, which is what you would need to get a 2-theory of HOL by lifting a 1-theory inside the 2-theory of FOL.

This notion of “lifting” (which was just a nonce word, by the way – if we were going to write down a serious theory of it we should think more carefully about what to call it) already appeared on this blog 10 years ago in the form of Tom’s “category-theoretic shibboleth”. The monoidal category freely generated by a monoid is (via the identification of theories with their initial algebras) the lifting of the 1-theory of monoid objects in the 2-theory of monoidal categories.

At the time, Tom described it as a categorification of the notion of a group presentation. So, for instance, the free group generated by a pair of elements $x,y$ such that $x y = y x$ is the lifting of the 0-theory of a pair of commuting elements in the 1-theory of groups.

Posted by: Mike Shulman on September 27, 2020 3:44 AM | Permalink | Reply to this

### Re: What is an n-theory?

I’m not sure whether you are talking at cross purposes – the Jacobs-Melham paper that David cited is about interpreting dependent type theory in classical higher order logic (as used in HOL4 and Isabelle/HOL) if I recall correctly. That’s a simple type theory with axioms (again, iirc).

Posted by: Sam Staton on September 28, 2020 8:29 AM | Permalink | Reply to this

### Re: What is an n-theory?

What’s different about the classical case?

I see how you can formulate HOL (intuitionistic or classical) by defining the propositions to be terms belonging to the subobject classifier (which would be bool in the classical case). But I’m saying that’s modifying the 2-theory directly, rather than lifting something inside the 2-theory of FOL as David suggested.

Posted by: Mike Shulman on September 28, 2020 6:11 PM | Permalink | Reply to this

### Re: What is an n-theory?

Thank you, I understand what you are saying now. (So perhaps it was me that was at cross-purposes…)

Posted by: Sam Staton on September 28, 2020 9:13 PM | Permalink | Reply to this

### Re: What is an n-theory?

… except I didn’t refresh and missed David’s interesting comment.

Posted by: Sam Staton on September 28, 2020 9:17 PM | Permalink | Reply to this

### Re: What is an n-theory?

I wasn’t referring back to the May 24 comment with the question of lifts, just trying to make better sense of where HOL fits in.

I was thinking about how one might explain the change in expressivity brought about by DTT to an analytic philosopher brought up on FOL, set theory, and perhaps a glimpse of HOL (second-order logic, at least). Something like:

With the new possibilities opened up by the richer judgment structure (3-theory) of DTT, the pressure for a 1-theory of sets written in the 2-theory FOL is removed. Upgrading to a 2-theory HOL in that same 3-theory can be seen as a fix (giving rise to Quine’s comment on second-order logic as “set theory in sheep’s clothing”), but now with the proper resources of DTT, we can define powerful 2-theories such as HoTT.

Posted by: David Corfield on September 28, 2020 8:22 PM | Permalink | Reply to this

### Re: What is an n-theory?

If you find such a philosopher who’s willing to listen when you talk about “3-theories”, you can try it out! (-:

Posted by: Mike Shulman on September 28, 2020 10:10 PM | Permalink | Reply to this

### Re: What is an n-theory?

I’ve been lurking while you brought up encoding dependent type theory into HOL. It looks like this encoding is similar to things I know about.

David wrote:

I was thinking about how one might explain the change in expressivity brought about by DTT to an analytic philosopher brought up on FOL, set theory, and perhaps a glimpse of HOL (second-order logic, at least).

Maybe you can practice on me. I’m not an analytic philosopher, but I might be the most anti-typing type theorist you’ll meet. :)

There is an irony to encoding dependent type theory in HOL to get a richer 3-theory for expressing more 1- and 2-theories: the HOL being targeted is itself a 2-theory, not a 3-theory, so there’s a “dimension collapse” going on. If this is acceptable, why bother with any of this n-stuff at all?

(I gather it has something to do with model theory, but the onus is on you to explain why this sort of model theory matters. I know that at the n-Cat Cafe, n-stuff is assumed to matter. Please explain why, anyway.)

Posted by: Matt Oliveri on September 29, 2020 3:02 AM | Permalink | Reply to this

### Re: What is an n-theory?

There is an irony…

I’m not sure I see what you mean. The encoding of some dependent type 2-theory (written in the DTT 3-theory) into a 2-theory HOL (written in the FOL 3-theory) comes at the cost of introducing something ‘funny’ to that 3-theory.

(Which makes me wonder whether the intensional Martin-Löf type theorist might think that defining a univalent universe as part of the HoTT 2-theory is similarly ‘funny’ in the context of DTT 3-theory.)

But thinking about levels, I wonder if there’s something to the thought that locating inferential power at a higher level is in a sense preferable, rather than, e.g., having to make your 1-theory specifications do the work as first-order set theory does. I guess the higher the level of the syntactic theory, the higher the level of the collection of models.

But presumably we don’t lift for the sake of it. Is there any point to $Lift(T^1_{ZFC})$ as a FOL 2-theory?

Posted by: David Corfield on September 29, 2020 10:28 AM | Permalink | Reply to this

### Re: What is an n-theory?

The encoding of some dependent type 2-theory (written in the DTT 3-theory) into a 2-theory HOL (written in the FOL 3-theory) comes at the cost of introducing something ‘funny’ to that 3-theory.

Oh I see, you’re thinking of it as converting a 2-theory from one 3-theory to another, so the level isn’t changed. (Isn’t the “FOL 3-theory” actually for any simply-typed predicate logic?) What is the “funny” thing introduced by this encoding? (I have a guess what you mean, but I should probably let you express it your way.)

(Which makes me wonder whether the intensional Martin-Löf type theorist might think that defining a univalent universe as part of the HoTT 2-theory is similarly ‘funny’ in the context of DTT 3-theory.)

I don’t know. Some of them seem to really like it though, funny or not. What would be funny about it?

But thinking about levels, I wonder if there’s something to the thought that locating inferential power at a higher level is in a sense preferable, rather than, e.g., having to make your 1-theory specifications do the work as first-order set theory does.

You would have to explain why it’s preferable, and in what sense. Also, what do you mean by inferential power? Logical strength? (Like consistency strength, proof-theoretic strength.)

I guess the higher the level of the syntactic theory, the higher the level of the collection of models.

Didn’t Mike confirm that somewhere in these comments?

But presumably we don’t lift for the sake of it. Is there any point to $Lift(T^1_{ZFC})$ as a FOL 2-theory?

Sorry, I didn’t understand any of the “lift” stuff.

Posted by: Matt Oliveri on September 30, 2020 3:23 AM | Permalink | Reply to this

### Re: What is an n-theory?

I should preface any answers with the comment that all I know of this $n$-theory approach comes from Mike, as outlined in this post and thread and his HoTTEST talk. My interest of late comes in wondering whether it can provide resources to think through what’s going on with the proposal to take HoTT as a foundational language for mathematics instead of set theory.

Whenever I use the word ‘funny’ I’m referring back to Mike’s comment:

Higher-order logic is a funny beast from a doctrinal perspective.

I took this to mean that having specified our 3-theory which governs the dependency structure and arity of judgments, that when we specify a 2-theory within it, i.e., the rules of the deductive system in which relevant 1-theories will be given, that the specification for HOL of a universe of propositions via rules is atypical.

Perhaps Mike has something helpful on what that might mean, in which case it would say something about the provision of rules for a univalent universe.

I’m not quite sure where I’m going with the ‘higher level is preferable’ thought. As systems in which do mathematics, it could be said that $Th^1_{ZFC}$ has devolved expressive power to a lower level (the 1-level) by comparison to $Th^2_{HoTT}$, but I’m not sure what that tells us.

All I know about ‘lifting’ comes from this thread. Here’s Mike’s first mention in this comment.

I’ve no idea whether Mike is thinking of pursuing the $n$-theory framework, but I was just looking to see whether I could use it for some of my own ends.

Posted by: David Corfield on September 30, 2020 12:12 PM | Permalink | Reply to this

### Re: What is an n-theory?

Thanks for these pointers.

Whenever I use the word ‘funny’ I’m referring back to Mike’s comment:

Higher-order logic is a funny beast from a doctrinal perspective.

I took this to mean that having specified our 3-theory which governs the dependency structure and arity of judgments, that when we specify a 2-theory within it, i.e., the rules of the deductive system in which relevant 1-theories will be given, that the specification for HOL of a universe of propositions via rules is atypical.

Hmm. Well it is true that HOL is usually specified as having a Russell-style universe of propositions, not Tarski-style. Otherwise it seems pretty normal. We’ll see what he says.

I’m not quite sure where I’m going with the ‘higher level is preferable’ thought. As systems in which do mathematics, it could be said that $Th^1_{ZFC}$ has devolved expressive power to a lower level (the 1-level) by comparison to $Th^2_{HoTT}$, but I’m not sure what that tells us.

My understanding is that different models of a 1-theory within a model of the 2-theory would all be sharing the same interpretation of the stuff from the 2-theory. So more stuff in the 2-theory is more stuff that 1-theories are “taking for granted”. Traditional foundationalists seem extremely confident in the choice of classical first-order logic (a particular 2-theory) as giving exactly those things it’s safe to take for granted by a foundation.

I’m not so traditional, but I find it odd to regard HoTT as a 2-theory, especially if it has strong principles like induction. This is not particular to HoTT: dependent type theorists following Martin-Löf tend to combine the logic, proof system, and foundational theory all into one rather complicated thing. HoTT formulated as a dependent type system is just an example of this.

I prefer to regard these rich type systems as convenience layers for some more traditionally-organized foundational theory. I’m not sure how this should be done for HoTT, since I don’t really understand the models, but I still think it might be good to do it somehow, or say why it’s impossible.

Posted by: Matt Oliveri on September 30, 2020 3:03 PM | Permalink | Reply to this

### Re: What is an n-theory?

I think y’all are reading too much into my “funny” comment. It was made over 2 years ago and I’m no longer entirely sure what I meant. I suspect I was just referring to the fact that, as David had just suggested, at first it seems reasonable to expect HOL to be a 3-theory, whereas in fact its dependency and arity structure is the same as that of FOL so they should really share a 3-theory (which means that, as I pointed out in the same comment, “FOL” is indeed not a good name for that 3-theory).

In other words, it’s not so much that HOL is intrinsically funny, but that the $n$-theories perspective leads us to a different view of its relationship to FOL and DTT: instead of placing the three on an equal footing as foundational choices to be compared, FOL and HOL are placed in one group that is distinguished much more sharply from DTT than from each other. Kind of like how biologists sometimes have to reshuffle their division of species into genera and families upon learning from genetic evidence that different-looking species are actually closely related.

I don’t think I have an immediate reaction to the question of whether it’s “better” to locate “inferential power” at a higher or lower level. My instinct is that this is a question on which we should be descriptive rather than prescriptive. ZFC does it one way and MLTT/HoTT does it another way, but each choice seems to work pretty well for them.

Posted by: Mike Shulman on September 30, 2020 4:44 PM | Permalink | Reply to this

### Re: What is an n-theory?

One feature about having the inferential power at a lower level is that when one wants to use it to reason about some aspect of the world, there’s no separate level to specify worldly elements. Hence from the 1-theory ZF, one forms a 1-theory of sets and ‘atoms’ or ‘ur-elements’ in the form of ZFA.

Starting from, say, a dependent type 2-theory, one may form a worldly 1-theory by specifying some worldly types.

Posted by: David Corfield on October 20, 2020 10:25 AM | Permalink | Reply to this

### Re: What is an n-theory?

Come to think of it, maybe this difference in style between ZFA and dependent type theory has nothing to do with ZFA being a 1-theory.

What about ETCS and SEAR? These are (possibly dependently-sorted) first-order theories… Aren’t they 1-theories? But I figure specifying a “worldy” theory in ETCS or SEAR would be like in dependent type theory, except with different syntax.

Maybe ETCS and SEAR should be considered 2-theories, as if they are dependent type systems with funny syntax. But then that would make the outer predicate logic into a 3-theory, as if it were a logical framework or something.

I don’t know. I thought the assignment of theories to levels is based on the dimension of the intended class of models. But I figure this ought to be independent of syntactic issues. So this would make it seem like ETCS and a dependent type system should have the same level.

On the other hand, Mike has been saying things making it sound like the different levels are responsible for specific aspects of the syntax, if I understood correctly. So then it might be more significant that ETCS is a first-order theory.

Getting back to ZFA, I think the elephant in the room is that material set theory totally punts on handling higher dimensions, so it’s a 1-theory by default. But it’s even “worse” because it’s untyped (in a sense). This all seems like a deliberate simplification. (Well, deliberate for set theorists; I don’t know about the philosophers trying to make direct use of minor variants of material set theory.) If material set theory’s approach to modeling worldly issues is considered unsatisfactory, I’m not convinced this is because it’s a 1-theory rather than simply because it’s untyped.

Posted by: Matt Oliveri on October 25, 2020 6:42 PM | Permalink | Reply to this

### Re: What is an n-theory?

What about ETCS and SEAR? These are (possibly dependently-sorted) first-order theories… Aren’t they 1-theories?

Yes.

Posted by: Mike Shulman on October 25, 2020 6:52 PM | Permalink | Reply to this

### Re: What is an n-theory?

Is there some natural way to regard them as 2-theories instead? The intuition being to treat the sets as types?

Posted by: Matt Oliveri on October 25, 2020 9:37 PM | Permalink | Reply to this

### Re: What is an n-theory?

You could certainly write down 1-theories in the 2-theory of simple type theory, or the 2-theory of dependent type theory, that are equivalent in some sense to ETCS or SEAR. But one might dispute whether those are the same theory. After all, ZFC is equivalent to ETCS+R in an appropriate sense, but we still regard them as different theories.

Posted by: Mike Shulman on October 25, 2020 10:20 PM | Permalink | Reply to this

### Re: What is an n-theory?

Let me try asking basically the same question a different way:

What is a theory whose models are structured categories that have all the structure and properties of the category of sets that are visible to ETCS?

Posted by: Matt Oliveri on October 25, 2020 10:57 PM | Permalink | Reply to this

### Re: What is an n-theory?

I don’t know of a theory that has exactly those categories as models, but if you rearrange the structure a bit, into a hyperdoctrine or CwF, then HOL or some version of MLTT should work respectively.

Posted by: Mike Shulman on October 26, 2020 12:09 AM | Permalink | Reply to this

### Re: What is an n-theory?

I’m confused. Doesn’t every set-theoretic model of ETCS induce such a category? (And all such categories arise that way?) Why can’t you just declare that to be the notion of model you’re interested in?

Maybe I need more details on how you would “rearrange the structure a bit”. I don’t understand why it needs rearranging.

Posted by: Matt Oliveri on October 26, 2020 3:57 PM | Permalink | Reply to this

### Re: What is an n-theory?

I thought you were looking for an answer that was a type theory whose corresponding notion of structured category was an ETCS-category, in the same way that STLC corresponds to CCCs and so on. To do that I think you need to draw out some of the properties of an ETCS-category into explicit structure such as a hyperdoctrine or CwF, as I said. Did you instead want me to give the trivial answer “ETCS”?

Posted by: Mike Shulman on October 26, 2020 4:01 PM | Permalink | Reply to this

### Re: What is an n-theory?

Yeah, kind of! This started with me wondering if ETCS and SEAR could also be regarded as 2-theories. My understanding (possibly all wrong, which would explain how this subthread got off course) is that if you can use ETCS as a theory whose models are structured categories, you’re using it as a 2-theory.

I thought you were looking for an answer that was a type theory whose corresponding notion of structured category was an ETCS-category, in the same way that STLC corresponds to CCCs and so on.

Yes. But here, I’m not being fussy about syntax. So ETCS counts as a type theory, as long as there’s no technical problem interpreting it that way.

Getting something that looks more like Martin-Löf type theory might be nice. But I figured this should not be necessary.

To do that I think you need to draw out some of the properties of an ETCS-category into explicit structure such as a hyperdoctrine or CwF, as I said.

My understanding is that we’d need a notion of functor between ETCS-categories, to have the 2-category of models. But I was guessing people knew how to get a good notion of functor in this case without making the structure more explicit.

And I guess we’d want an initiality theorem. Is there a known one?

Posted by: Matt Oliveri on October 26, 2020 7:09 PM | Permalink | Reply to this

### Re: What is an n-theory?

My understanding (possibly all wrong, which would explain how this subthread got off course) is that if you can use ETCS as a theory whose models are structured categories, you’re using it as a 2-theory.

The dimension of a theory is not defined in terms of the apparent categorical dimension of its models. When ETCS is formulated as usual in the language of (possibly dependently-typed) first-order logic, it is a 1-theory because it plays the same syntactic role as ZFC or the theory of a group do inside ordinary one-sorted first-order logic: it specifies some sorts, function-symbols, relation-symbols, and axioms. The fact that its models “are” categories is of no significance to the syntactic role of the theory. Similarly, but more simply, the essentially algebraic theory of categories is also a 1-theory. A 2-theory is something that specifies the type- and term-forming operations.

To put it more semantically, there is both a 1-category of categories (or ETCS-categories) and a 2-category of categories, and hence both a 1-theory of categories and a 2-theory of categories. Arbitrary first-order logic doesn’t “know” how to define “2-morphisms” between homomorphisms of its models (and in most cases there is no meaningful such notion). Thus when a theory of categories such as ETCS is formulated in first-order fashion, it can only correspond to the 1-category of models and hence be a 1-theory.

Posted by: Mike Shulman on October 26, 2020 7:34 PM | Permalink | Reply to this

### Re: What is an n-theory?

Similarly, but more simply, the essentially algebraic theory of categories is also a 1-theory.

I guess it would be easier to get on the same page with this simpler case.

A 2-theory is something that specifies the type- and term-forming operations.

I would like to regard the essentially algebraic theory of categories as specifying these things: objects are types and morphisms are terms.

Both of these would normally be considered terms, syntactically. Is that really what determines it, in your classification into levels?

To put it more semantically, there is both a 1-category of categories (or ETCS-categories) and a 2-category of categories, and hence both a 1-theory of categories and a 2-theory of categories.

OK. What 2-theory of categories should we discuss?

Posted by: Matt Oliveri on October 27, 2020 2:45 AM | Permalink | Reply to this

### Re: What is an n-theory?

The dimension of a theory is not defined in terms of the apparent categorical dimension of its models.

This is very helpful to me, except for that “apparent” in there, making me think I’m still missing something. Suppose we’re writing interpretation functions in HoTT. Surely there is no ambiguity what the dimension of the type of models is.

I understand that a generic interpretation function for FOL theories would result in a 1-type of models. But in the particular case of ETCS, this outcome kind of looks like an unfortunate accident. So I’m looking for a different notion of model for the same syntax, by adjusting the interpretation.

Supposing this adjusted interpretation is used, and consequently the type of models is a 2-type, I don’t appreciate why ETCS should not then be considered a 2-theory.

A 2-theory is something that specifies the type- and term-forming operations.

But given the intent of regarding ETCS’ sets as the types, there’s a sense in which ETCS does these things, just with unusual syntax.

Arbitrary first-order logic doesn’t “know” how to define “2-morphisms” between homomorphisms of its models (and in most cases there is no meaningful such notion).

OK. If I understand correctly, this screws up the possibility of regarding FOL as a 3-theory.

Thus when a theory of categories such as ETCS is formulated in first-order fashion, it can only correspond to the 1-category of models and hence be a 1-theory.

Here I’m not convinced. It seems like you’re conflating the way a theory is formulated (syntax) and interpreted (semantics). If FOL is not a 3-theory, then ETCS will need some novel semantics to be a 2-theory. (And there is an implicitly “missing” 3-theory saying how to generalize this.) But you haven’t ruled out the possibility, as far as I can tell.

Posted by: Matt Oliveri on October 27, 2020 2:27 AM | Permalink | Reply to this

### Re: What is an n-theory?

there’s a sense in which ETCS does these things, just with unusual syntax.

Yes, but that isn’t the sense that is relevant for deciding whether it is a 2-theory (in the way I’m using the word). As I said before, there’s a strong sense in which ETCS+R and ZFC are equivalent, but we still distinguish between them. Similarly, there’s a sense in which ETCS is equivalent to a 2-theory, but that doesn’t make it one.

ETCS will need some novel semantics to be a 2-theory.

The moment you give it “novel semantics” you’re not talking about ETCS any more. Just as an object is always an object of some category, a theory is always a theory in some doctrine. And just as the morphisms into and out of of that object are determined by the category in which it lives, the semantics of a theory are determined by the doctrine in which it lives.

Posted by: Mike Shulman on October 27, 2020 3:59 AM | Permalink | Reply to this

### Re: What is an n-theory?

Maybe I’m phrasing things too dogmatically. It’s certainly true that there is a 2-category whose objects are models of ETCS, and so one might argue from a semantic perspective that that can be regarded as a 2-theory, just like the 2-category of finite-limit categories is the semantic version of the 2-theory of essentially algebraic theories. I think what I’m saying is just that I wouldn’t call that 2-theory “ETCS”, maybe say “$ETCS_2$” or something. For one thing, I don’t know in what 3-theory that 2-theory would live; for another (relatedly), I don’t know what its syntax would look like.

Posted by: Mike Shulman on October 27, 2020 4:20 AM | Permalink | Reply to this

### Re: What is an n-theory?

there’s a strong sense in which ETCS+R and ZFC are equivalent, but we still distinguish between them. Similarly, there’s a sense in which ETCS is equivalent to a 2-theory, but that doesn’t make it one.

Well OK, but I thought the equivalence between 1-theory and 2-theory versions would be a lot less work.

The moment you give it “novel semantics” you’re not talking about ETCS any more. Just as an object is always an object of some category, a theory is always a theory in some doctrine.

So it sounds like you want to consider a theory as intrinsically including enough of its interpretation to identify a notion of model. I was just thinking of a theory as pure syntax.

It’s certainly true that there is a 2-category whose objects are models of ETCS, … just that I wouldn’t call that 2-theory “ETCS”, maybe say “$ETCS_2$” or something. For one thing, I don’t know in what 3-theory that 2-theory would live; for another (relatedly), I don’t know what its syntax would look like.

Well I don’t know what 4-theory any of these 3-theories live in. I figured at some point, you just work out the interpretation the old fashioned way. I was proposing to do that for “$ETCS_2$”, starting from the 1-interpretation (?) that comes out of FOL.

As for the syntax, I wanted it to be exactly the same. (That’s what made it the same theory, for me.)

BTW, I’m not serious about working this out in detail. This was a thought experiment to understand how levels are assigned to theories.

Posted by: Matt Oliveri on October 27, 2020 4:56 PM | Permalink | Reply to this

### Re: What is an n-theory?

Just like theories have both a syntactic and a semantic version, doctrines (i.e. 2-theories) have both syntactic and semantic versions. The syntactic version of the doctrine specifies the judgment forms of a theory. So you can’t make ETCS into a 2-theory and leave its syntax exactly the same, because the structure of the syntax is that of a theory in the doctrine of first-order logic.

Posted by: Mike Shulman on October 27, 2020 5:02 PM | Permalink | Reply to this

### Re: What is an n-theory?

Hmm, OK. So ETCS and $ETCS_2$ can’t have the same syntax.

So n-theories have a mandatory syntactic component. Why is it done that way? It seems like you would very soon run out of important distinct aspects of syntax to control with higher n-theories. Why not use one “regime” (for lack of a better word) to organize syntax, and save the regime of n-theories to organize the semantics?

Posted by: Matt Oliveri on October 27, 2020 9:08 PM | Permalink | Reply to this

### Re: What is an n-theory?

The whole point of the notion (indeed, one might say the whole point of logic) is that there is a correspondence between syntax and semantics. Divorcing them would vitiate the purpose.

Posted by: Mike Shulman on October 27, 2020 9:23 PM | Permalink | Reply to this

### Re: What is an n-theory?

Using different “regimes” for syntax and semantics doesn’t eliminate all correspondence between syntax and semantics. It just means the interpretation of syntax might need to do something creative to mediate between regimes.

Trying to use the exact same syntax for ETCS and $ETCS_2$, or for the 1-category and 2-category of categories, are maybe not sufficient motivation to engineer such a regime change.

But isn’t something more flexible than n-theories going on in the interpretation of HoTT? We want to think of the models as (∞,1)-categories, but the syntax doesn’t seem to have anything remarkable enough to be a (∞,2)-theory, I figure. The interpretation that yields a (∞,2)-category of models (or does it?) involves forgetting presentation details.

Well why can’t I also forget presentation details in ETCS models to get the intended categories as models?

For another example, how about bidirectional vs unidirectional type systems? The judgment structure is different, but I figure the intended collection of models has nothing to do with this difference.

My concern is that there are various engineering reasons to vary the syntax and semantics independently, and this n-theories approach as you’ve explained it seems to not allow it.

Posted by: Matt Oliveri on October 27, 2020 11:29 PM | Permalink | Reply to this

### Re: What is an n-theory?

As I said in the main post,

Homotopy type theory is a good example: despite being ostensibly a theory of $(\infty,1)$-categories, as a type theory it is actually just a 2-theory, or a deductive system for a certain class of 1-theories. Syntactically, this is roughly because it doesn’t include any basic 2-cell judgments; its 2-cells and higher cells are all encoded implicitly using 1-morphisms into path objects. Semantically, this corresponds to the fact that its most “direct” notion of model is actually a kind of structured 1-category (a comprehension category, category with families, etc.); a coherence theorem then (hopefully) relates these to the intended $(\infty,1)$-categorical models.

You’re right that ETCS is an instance of this same phenomenon in lower dimensions: the intended models are 1-categories, but since ETCS is a 1-theory its most direct notion of model is actually a kind of structured 0-category. But in my terminology, that doesn’t change the fact that HoTT is a 2-theory and ETCS is a 1-theory.

The framework I’m describing is a bridge between syntax and semantics, pairing up a kind of syntax and a kind of semantics that match up very closely, with the goal of systematizing the syntax-semantics correspondence. Certainly there are also reasons to allow more variability on both sides: as you point out, one can directionalize the judgments in syntax, and one can apply coherence theorems to homotopify the models in semantics. But the suggestion is that it’s easier to relate syntax to semantics by building a bridge between pairs that match closely, and then when thinking about bidirectionalism or coherence theorems you can stay completely in the worlds of syntax or semantics, respectively.

Posted by: Mike Shulman on October 28, 2020 1:58 AM | Permalink | Reply to this

### Re: What is an n-theory?

OK, that sounds much better. And I did remember that providing syntactic systems closely corresponding to categorical semantics was a goal of the LSR program. But using the term “n-theory” just according to this close correspondence doesn’t seem natural to me.

I was going to tell you why, but you convinced me to reread the main post, and then I got confused. The main post says:

The models of an $(n+1)$-theory in the $(n+1)$-category $n Cat$ are the semantic $n$-theories in that $(n+1)$-theory.

Now set ($n \coloneqq m + 1$). The models of the $(m+2)$-theory in $(m+1)Cat$ are given by functors. So these models are generally diagrams of $(m+1)$-categories. (As you pointed out much later in the post. (I think that’s what you meant.)) But you say they are $(m+1)$-theories. So then we should consider functors to $m Cat$ to find the $m$-theories. But what is a functor from a diagram of $(m+1)$-categories to the single $(m+1)$-category $m Cat$?

Maybe I’m supposed to figure that $n$-doctrines are the one-object special case of $(n+1)$-theories? (So that the models are structured $n$-categories, not diagrams.) And that only $n$-doctrines provide a notion of $n$-theory?

In other words, we have:

• A semantic $n$-theory is a structured $n$-category.
• Models of an $n$-theory are functors from its structured $n$-category to another structured $n$-category.
• A semantic $n$-doctrine is a semantic $(n+1)$-theory with one object.
• A semantic $n$-theory of some $n$-doctrine is a model of that doctrine in $n Cat$.
Posted by: Matt Oliveri on October 29, 2020 4:49 AM | Permalink | Reply to this

### Re: What is an n-theory?

Yes, that’s a good point. Moreover, single-sortedness of an $(n+1)$-theory isn’t enough to guarantee that $n Cat$ is a model of it at all — or, if it is, that there is a unique choice of such structure. For instance, $Set$ ($= 0Cat$) models the 1-theory of symmetric monoidal categories in the obvious way with its cartesian product, but also with its cartesian coproduct, and it doesn’t model the 1-theory of $\ast$-autonomous categories at all. So for full generality we should incorporate the choice of an “$n Cat$-like model” at each dimension. I’m not sure if there’s a clever way to state that, though; I tried for a few minutes but there’s an interesting sort of “turtles all the way up” problem.

Posted by: Mike Shulman on October 29, 2020 4:48 PM | Permalink | Reply to this

### Re: What is an n-theory?

I was wondering back here about the requirement to model in $Set (= 0 Cat)$, since the authors mentioned were looking to $Rel$.

Posted by: David Corfield on October 31, 2020 4:42 PM | Permalink | Reply to this

### Re: What is an n-theory?

I think I see what you mean. I found one reason why things need to be more degenerate/normal/pure as you go up, but there are others.

Ignoring the problem of singling out $n Cat$ and the maps into it, I thought of an improvement to my fix proposal yesterday. In that proposal, semantic n-theories had to be structured n-categories, but it seems likely that that messes up LSR’s support for modes. Instead:

• A semantic $n$-doctrine is a structured $(n+1)$-category.
• Models of an $n$-doctrine are functors from its structured $(n+1)$-category to another structured $(n+1)$-category.
• A semantic $n$-theory of some $n$-doctrine is a model of that doctrine in $n Cat$. (A functor, and a diagram of $n$-categories.)
• Models of an $n$-theory are natural transformations from its semantic theory to another semantic $n$-theory.

As for what I called a doctrine yesterday, I can make up a tacky new name for it, like “doctriverse”. It seems that theories are like elements, and doctrines are like types. (You hinted at this.) So we have doctrines whose elements are doctrines, like types whose elements are types, which are called universes. Hence “doctriverses”.

What about doctrines whose elements are doctriverses? I don’t see why not.

ZFA (ironically) seems to shed some light on this degeneracy/purity issue. An object of the theory might be an atom, meaning it’s not a set. Only membership in sets can be nontrivial. (And set extensionality only applies to sets, but that aspect doesn’t fit here.)

Objects (of ZFA) correspond to theories; sets correspond to doctrines. (Atoms, or “urelements”, correspond to theories that aren’t doctrines.) There’s a ranking of ZFA objects that works like rank of pure sets and basically ignores atoms. So you can have a set of any rank, which may or may not have atoms. But you could also define a (cumulative) purity degree:

• Every object has purity degree 0.
• A set has purity degree (n+1) if all its elements have degree n.

So a pure set has all (natural number) purity degrees, an atom only has 0, a set has at least up to 1. (I don’t think transfinite degrees make sense.)

If I haven’t screwed something up, the idea is that theories have to have purity degree 0, docrines have to have 1, doctriverses have to have 2, etc. A purity degree of n guarantees you have a notion of nested theories for n levels.

The dimension $n$ in front of “$n$-theory” seems more like a rank than a purity degree, but it’s certainly not a material set rank. The point is that the dimension and purity degree seem pretty independent.

I think this is pretty interesting, anyway.

The problem can now be stated as: What crazy thing is a semantic $n$-theory in general, and how do you determine its purity degree?

Posted by: Matt Oliveri on October 29, 2020 9:21 PM | Permalink | Reply to this

### Re: What is an n-theory?

If you stop at the $(n+k)^{\mathrm{th}}$ turtle going up, then an “$n$-theory” below it has “purity degree” $k$. Perhaps instead of trying to define “a semantic $n$-theory” in general and then afterwards asking what its purity degree is, we have to define “a semantic $n$-theory of purity degree $k$” by induction on $k$.

This kind of reminds me of the putative example of an “$\omega^{op}$-category” — like an $\omega$-category, but with cell dimensions indexed by nonpositive integers rather than nonnegative ones — that Peter Lumsdaine once suggested. Start with the 0-cells being sets. But a set is just a functor $1\to Set$, so we can take the $(-1)$-cells to be categories and expand the 0-cells to be functors between them, including our previous 0-cells as $Hom(1,Set)$. But now a category is just a 2-functor $1\to 2Cat$, so we can take the $(-2)$-cells to be 2-categories and expand the $(-1)$-cells to be 2-functors between them, including our previous $(-1)$-cells as $Hom(1,Cat)$. And “so on”, whatever that means. (We can also go up one more dimension and take the 1-cells to include elements of sets, and perhaps we ought to reindex everything so that those are the 0-cells.)

Posted by: Mike Shulman on October 30, 2020 4:57 AM | Permalink | Reply to this

### Re: What is an n-theory?

Perhaps instead of trying to define “a semantic $n$-theory” in general and then afterwards asking what its purity degree is, we have to define “a semantic $n$-theory of purity degree $k$” by induction on $k$.

What’s the difference? Degree 0 is the general case.

Do you mean—since we’re stopping (on the way up) at dimension $(n+k)$—that we do induction counting down from $(n+k)$? This sounds more like what ought to happen, since the possible theories seem to be getting more complicated on the way down.

What are $n$ and $k$ in the LSR program?

This kind of reminds me of the putative example of an “$\omega^{op}$-category”…

I had a weird idea quite a while ago, based on the intuition that dimensions are kind of like set ranks, and so it might be interesting to try and make them even more like ranks.

Doing infinity categories based on presheaves like simplicial/cubical sets seems like trying to explain nonwellfounded sets. But the history of set theory seems to suggest that that’s doing it the hard way.

I was wondering if there could be a theory of things like ∞-groupoids (like HoTT), except

• All spaces are truncated.
• But the truncation level can be any ordinal.

So this would be a theory of $\alpha$-groupoids. There would be “higher” cells for all natural numbers, but things are well founded in that a cell of a $\alpha$-groupoid will be some gadget of dimension $\beta$ where $\beta \lt \alpha$. (And that ordinals are wellfounded, of course.)

This was only intended to try and get a different explanation of higher equality, so it doesn’t directly address the problems of general $n$-theories. But I have some vague, possibly-relevant thoughts from this:

• Maybe it’s annoying that they’re called “higher” cells when they form collections of lower dimension. It makes Lumsdaine’s thing look upside down. Maybe it’s topology that’s upside down.

• Maybe it’s enough to say what a theory is by counting down finitely in arbitrarily complex ways. (Decreasing sequences of ordinals)

• Maybe it’s not so crazy to look for a notion of totally pure doctrine, like pure sets. Similarly, why can’t every HoTT type be a univalent universe?

(This was obviously an evil plan to turn HoTT into material set theory, but with higher equality. At least there would be no negative numbers. :))

This idea is very incomplete; it could be full of holes. But maybe someone will enjoy thinking about it. (I personally will probably not try to work out any new higher anything in detail until after I work out some other things.)

Posted by: Matt Oliveri on October 31, 2020 4:05 AM | Permalink | Reply to this

### Re: What is an n-theory?

Yes, I meant down from $n+k$. Maybe I misunderstood what you meant by “purity degree”; I meant that at dimension $n+k$ you require models in $(n+k-1)Cat$ and then go down to define what an “$n$-theory” is relative to this choice of $k$. I think LSR have so far only done $k=1$, where the 2-theory takes models in $Cat$.

Posted by: Mike Shulman on November 1, 2020 8:46 AM | Permalink | Reply to this

### Re: What is an n-theory?

I meant that at dimension $n+k$ you require models in $(n+k-1)Cat$ and then go down to define what an “$n$-theory” is relative to this choice of $k$.

Right. The observation was that not all $(n+k)$-theories provide a sensible way of going down $k$ levels. (Unless I completely misunderstood.) It has to be sufficiently “pure”. But I sure don’t know what that actually entails.

I think LSR have so far only done $k=1$

Wait so the LSR program has type theories that are 1-theories, and “mode theories” that are 2-theories? What about the language in which the mode theories are written? That’s the 3-theory? Doesn’t it have purity degree 2, somehow? So I was guessing it was at least similar to $n = 1$, $k = 2$, with the semantics of the top level done manually.

I was wondering: If there are axioms (propositional hypotheses) for the 1-theory, does that count as a 0-theory? Or is that just part of the 1-theory in some other way? I don’t just mean in the case of LSR; so far I didn’t think it had axioms. I mean a 1-theory in FOL, for example. Does the difference between signature and theory correspond to a division between 1- and 0-theories?

Posted by: Matt Oliveri on November 1, 2020 6:00 PM | Permalink | Reply to this

### Re: What is an n-theory?

Well, I don’t really know what you mean by “purity degree”. You can start indexing $k$ at 0 or 1 as you choose.

Axioms in the 1-theory can either be regarded as a 0-theory written in the 1-theory, or as part of a “lifted” 1-theory, as discussed here and here.

Posted by: Mike Shulman on November 2, 2020 3:18 AM | Permalink | Reply to this

### Re: What is an n-theory?

OK, probably nothing to worry about. Thanks very much for trying to explain things.

Posted by: Matt Oliveri on November 2, 2020 5:51 AM | Permalink | Reply to this

### Re: What is an n-theory?

Getting back to this:

that sounds much better. And I did remember that providing syntactic systems closely corresponding to categorical semantics was a goal of the LSR program. But using the term “n-theory” just according to this close correspondence doesn’t seem natural to me.

I thought there should be a different, more specific term for (syntactic) n-theories whose syntax is organized in accordance with an LSR-like “regime”.

The main post says that:

a syntactic 2-theory will be a way of “presenting” such a structured 2-category in terms of generating objects, morphisms, 2-morphisms, and relations.

So I thought, tl;dr, a syntactic n-theory is a syntactic presentation of a semantic n-theory. But you’re saying now that $ETCS_2$ can’t have the same syntax as ETCS. Well the syntax of ETCS seems to function as a presentation of the semantic 2-theory $ETCS_2$. So it’s like you’re saying ETCS presents it wrongly.

The extra restrictions are what I’m calling the “regime”, since it sounds like the restriction is that aspects of the syntax are imposed top-down. This corresponds to the way down from (n+k) to n, I suppose: A syntactic (n+i+1)-theory is apparently not just a syntactic presentation of a semantic (n+i+1)-theory; it’s also a partial specification of how its semantic (n+i)-theories should be syntactically presented.

I was going to suggest that syntactic theories following this sort of regime be called “canonically presented” or something.

But I changed my mind. The syntactic regime seems like a generalization of the usual usage of predicate logic, which is where the term “theory” comes from.

So maybe an arbitrary syntactic presentation of a semantic n-theory should be called an “n-language”. Then you can take ETCS (a 1-theory), and consider the language as a whole (including FOL) as a 1-language in the usual way, or as a 2-language, by adjusting the interpretation. What do you think?

Posted by: Matt Oliveri on October 31, 2020 4:07 AM | Permalink | Reply to this

### Re: What is an n-theory?

Ok, sorry for putting too much pressure on your words. I’m trying to see how the arrangement of species looks in the new classification.

Two more to fit in

Makkai’s first-order logic with dependent sorts (FOLDS), and Aczel’s and Belo’s dependently typed (intuitionistic) first-order logic (DFOL)

mentioned in the abstract here.

Posted by: David Corfield on September 30, 2020 9:43 PM | Permalink | Reply to this

### Re: What is an n-theory?

Well, just as ordinary HOL and FOL belong to the same 3-theory, I would expect dependently sorted first-order logic to belong to the same 3-theory as other dependent type theories.

Posted by: Mike Shulman on October 2, 2020 3:38 PM | Permalink | Reply to this

### Re: What is an n-theory?

Returning to this:

The encoding of some dependent type 2-theory (written in the DTT 3-theory) into a 2-theory HOL (written in the FOL 3-theory) comes at the cost of introducing something ‘funny’ to that 3-theory.

It sounds like the encoding is not “funny” then?

So to play devil’s advocate: If it’s possible to convert dependent type theories into versions of predicate logic, why should model theorists care about dependent type theory? It sounds like syntax tricks. Is the “change in expressivity” you mentioned just an illusion?

But presumably we don’t lift for the sake of it. Is there any point to $Lift(T^1_{ZFC})$ as a FOL 2-theory?

Maybe models of that theory are Boolean-valued models of ZFC? (Nonstandard truth values) In that case, there is supposedly a point.

Posted by: Matt Oliveri on October 1, 2020 1:21 AM | Permalink | Reply to this

### Re: What is an n-theory?

If it’s possible to compile <insert your favorite programming language> to machine code, why should programmers care about the former?

Posted by: Mike Shulman on October 1, 2020 1:57 AM | Permalink | Reply to this

### Re: What is an n-theory?

If it’s possible to compile…

There are two problems if I interpret that literally.

1. I don’t have a favorite programming language.
2. I am cynical about programmers’ reasons for caring about specific technologies.

1. It’s easier to program in a high-level language.
2. Multiple compilers for the same language.

These would correspond roughly to:

1. DTT is easier to use than predicate logic.
2. DTT theories have more models than predicate logic theories.

(1) sounds right. (2) doesn’t.

Does the reduction of a DTT to a predicate logic lose models? I thought it did not, and that in fact the usual way to interpret dependent type theory into categories is effectively via a reduction to predicate logic. (Interpret the judgments as predicates on raw terms.)

Posted by: Matt Oliveri on October 1, 2020 8:32 AM | Permalink | Reply to this

### Re: What is an n-theory?

As a standard contemporary English rhetorical device, the phrase “your favorite” in this context means “an arbitrary”.

Does the reduction of a DTT to a predicate logic lose models?

Not as far as I know, in the current state of knowledge. Your (1) is all I had in mind.

Posted by: Mike Shulman on October 2, 2020 3:35 PM | Permalink | Reply to this

### Re: What Is an n-Theory?

I wonder what we can make of our efforts to give a type theoretic account of necessity and possibility. As you know this concerns the use of a type of worlds, $W$, in an $(\infinity, 1)$-topos, $\mathbf{H}$, and the adjoint triple between the slice $\mathbf{H}/W$ and $\mathbf{H}$, by pullback and right and left adjoints. Here we can talk about these two $(\infinity, 1)$-toposes via a modal dependent type theory.

So I guess these, and other triples generated by an $\mathbf{H}$-morphism $f: W \to V$, are all models of a simple modal 2-theory, where the 2-category of modes, $\mathcal{M}$, contains an adjunction (slide 20). It could be a ‘walking’ such 2-category. Then these slices would provide a ready collection of models.

Working backwards, one might discover possible worlds semantics. I want an adjunction between a ‘necessity’ comonad and a ‘possibility’ monad, so I’m looking for models of a modal type theory for a 2-category of modes with a monad. One place to find them is between slices of $(\infinity, 1)$-toposes. What to make of the slicing objects? Well objects are varying over them, let’s call them ‘worlds’.

But perhaps the semantic theory comprised of pairs of slices of $(\infty, 1)$-toposes can be characterised syntactically.

Posted by: David Corfield on May 5, 2018 10:38 AM | Permalink | Reply to this

### Re: What Is an n-Theory?

Continuing on the modal theme, you list some modalities, including those for “differential cohesion”, and then write

I fully expect that in the future such modal type theories will proliferate.

Of course, there’s quite a story as to why differential cohesion with the addition of the fermionic-bosonic modalities is special, how it all arises by a process of Aubhebung from the opposition between the $\emptyset$ comonad and the monad $\mathbf{1}$, etc.

Can all of these requirements be expressed in the 2-category of modes?

Posted by: David Corfield on May 5, 2018 12:10 PM | Permalink | Reply to this

### Re: What Is an n-Theory?

Yes, there is a 2-theory generated by an adjunction, which in the semantics becomes an adjoint triple if we have both $p^\ast$ and $p_\ast$ for each mode morphism. In this case, the left adjoint is not left exact, so we have to be careful with dependency and contexts, but there’ll be some kind of type theory for sure. Since the adjunction isn’t idempotent, the general framework will produce a type theory with infinitely many “context zones”, but you could try to simplify it in this case.

perhaps the semantic theory comprised of pairs of slices of $(\infty, 1)$-toposes can be characterised syntactically.

Interesting question; I don’t know a syntactic characterization of etale geometric morphisms.

I don’t really speak Aufhebung, so all I can say to your last question is maybe.

Posted by: Mike Shulman on May 8, 2018 11:41 PM | Permalink | Reply to this

### Re: What Is an n-Theory?

I guess one could match a type theory with a different modal counterpart, e.g., dependent type theory with adjoint modalities, the later expressed in a unary type 2-theory of modes. But is the idea that its natural modal counterpart is a full dependent type 2-theory, as suggested by talk of the microcosm principle?

Have you given anywhere examples of constructions we might want to make which use the full resources of dependent type 2-theory?

Avoiding the off-putting Hegelianese, the question is whether some modal dependent type theory is expressive enough to specify 11 modalities, starting out from something like an adjoint comonad-monad $\emptyset \dashv \mathbf{1}$. Then we need a minimal pair $\flat \dashv \sharp$, such that $\sharp \emptyset \simeq \emptyset$, and such that $\flat$ has a further left adjoint monad. This monad, call it shape, is such that the shape of a point is a point, and we may require a points-to-pieces transform. And so on, with all the demands up to the solid modalities.

Posted by: David Corfield on May 9, 2018 10:19 AM | Permalink | Reply to this

### Re: What Is an n-Theory?

I guess one could match a type theory with a different modal counterpart.

I’m not sure what that would mean. The way our syntactic type theory has type judgments parametrized by mode judgments, the mode theory has to have the same context structure as the type theory. So with a unary mode 2-theory, it wouldn’t make sense to have a simple or dependent type theory over it.

the question is whether some modal dependent type theory is expressive enough to specify 11 modalities

We can certainly express any number of modalities, be it 6, 11, or 40 billion, and any algebraic relations between them like $\flat \to ʃ$ or $\flat \subseteq \&$. But not a property like minimality.

Posted by: Mike Shulman on May 9, 2018 5:25 PM | Permalink | Reply to this

### Re: What Is an n-Theory?

It strikes me that it may be worth saying a bit about how our setup is related to logical frameworks.

Roughly, the idea of a logical framework is that by formulating an object theory in a weak enough metatheory, it becomes easier to represent or prove metatheorems about the object theory. For instance, by cutting the metatheory down to a dependent type theory with only $\Pi$-types (no inductive types), we can use higher-order abstract syntax to avoid dealing manually with variable binding in the object language.

This sort of logical framework is, by design, extremely general in what sorts of object theories it can represent. Thus, while the logical framework provides a convenient context in which to prove theorems like cut-elimination and normalization, it doesn’t guarantee their truth; you can represent all sorts of ill-behaved theories as well. One way to describe our approach is to impose serious restrictions on the logical framework such that all object languages will satisfy the desired semantic and syntactic metatheorems, which can then be proved once and for all.

The logical framework used by Twelf and its relatives is a dependent type theory; thus it corresponds roughly to our 3-theory of “dependent type theories”. But there could also be “unrestricted” logical frameworks corresponding to other 3-theories. For instance, while I don’t completely understand Isabelle, I believe that its base system “Isabelle/Pure” can roughly be described as an (unrestricted) logical framework for the 3-theory that I called “first-order logic”, with one simple type theory dependent on another simple type theory. Unlike Twelf, Isabelle is designed to be used as a practical proof assistant for working in its object languages (rather than for proving metatheorems about them). The object language that it’s most commonly used for is called Isabelle/HOL, which is I think an implementation of (classical) higher-order logic inside the 3-theory of “first-order logic” exactly as I suggested here.

Posted by: Mike Shulman on May 25, 2018 5:32 PM | Permalink | Reply to this

### Re: What Is an n-Theory?

Can you give us a sense of the form of these ‘serious restrictions’?

Posted by: David Corfield on May 27, 2018 8:38 AM | Permalink | Reply to this

### Re: What Is an n-Theory?

William Orton, President of Western Union, 1876:

This ‘telephone’ has too many shortcomings to be seriously considered as a means of communication.

Thomas Edison, inventor, 1889:

Fooling around with alternating current (AC) is just a waste of time. Nobody will use it, ever.

Thomas Watson, president of IBM, 1943:

I think there is a world market for maybe five computers.

Mike Shulman, homotopy type theorist, 2018:

So, while in principle one could go on to talk about 4-theories, $n$-theories, and $\infty$-theories, it’s not clear to me that anyone will.

Luckily you were wise enough to be more cautious than some. The mere fact that you said this is going to make some young whippersnappers try to prove you wrong!

Posted by: John Baez on May 27, 2018 1:32 AM | Permalink | Reply to this

### Re: What Is an n-Theory?

More power to them! In fact, as I mentioned in another comment, they’ve already gotten started: opetopic type theory could be argued to be a true $(\infty,\infty)$-theory. (Of course, Eric Finster isn’t really any younger than me, so if he counts as a “young whippersnapper” then so should I.)

Posted by: Mike Shulman on May 27, 2018 4:06 AM | Permalink | Reply to this

### Re: What Is an n-Theory?

The hard part is not being young: by now most people are young. The hard part is being a whippersnapper.

For me the fun would not be developing the general formalism, but finding a bunch of really compelling examples of $n$-theories for high $n$, and using them to think about math in new ways.

Posted by: John Baez on May 28, 2018 6:05 PM | Permalink | Reply to this

### Re: What Is an n-Theory?

Note that in order to prove me wrong, I think one will also have to show that it’s better to treat such theories syntactically as “honest” $n$-theories, rather than encoding them as lower-dimensional theories the way HoTT encodes $(\infty,1)$-theories as 1-theories.

Posted by: Mike Shulman on May 28, 2018 11:10 PM | Permalink | Reply to this

### Re: What Is an n-Theory?

Mike, you speak of models of 1-theories in the “canonical” 1-category $Set$, and then similarly for higher $n$. What then do you think of the proposal of

• Filippo Bonchi, Dusko Pavlovic, Pawel Sobocinski, Functorial Semantics for Relational Theories, (arXiv:1711.08699)

to generalize Lawvere theories with a type of theory whose models land in $Rel$?

Whereas the universe for models of Lawvere theories is the category of sets and functions, or more generally cartesian categories, Frobenius theories take their models in the category of sets and relations, or more generally in cartesian bicategories.

Posted by: David Corfield on January 16, 2019 10:14 AM | Permalink | Reply to this

### Re: What is an n-theory?

Mike, your recent A practical type theory for symmetric monoidal categories concerns the “classical simple type (2-)theories” that you mention in the post above?

Posted by: David Corfield on November 7, 2019 9:13 AM | Permalink | Reply to this

### Re: What is an n-theory?

Yes, in that it’s a simple 2-theory that allows multiple types in both antecedents and consequents. However, it differs from classical logic and classical linear logic in the type of cut/substitution rule: it allows composition along multiple types at once (corresponding to a prop), whereas the others allow composition only along one type at a time (corresponding to a polycategory). I could imagine a 3-theory that can only include one of these kinds of composition, but I could also imagine a 3-theory that includes both.

Posted by: Mike Shulman on November 7, 2019 6:18 PM | Permalink | Reply to this

### Re: What is an n-theory?

You write in the post that you’re unclear whether there could be a 3-theory subsuming both the 3-theory of simple type 2-theories and the 3-theory of dependent type 2-theories. That’s presumably where a dependent linear type theory should live.

That page reports choices still to be made:

One of the most important is a theory that includes both “linear types” and “nonlinear types”, where the linear types may be dependent on the nonlinear types. The nonlinear types might also be allowed to depend on each other, but in this theory the linear types are not allowed to depend on each other.

There are other approaches to dependent linear type theory that do allow some sort of dependence between linear types, but we will not (yet) discuss them on this page.

To this there is now added the issue of composition along one or multiple types.

Posted by: David Corfield on November 9, 2019 8:03 AM | Permalink | Reply to this

### Re: What is an n-theory?

What I meant to say is that it’s unclear whether there’s a 3-theory that subsumes both the 3-theory(ies) of these “classical” simple type 2-theories (those with multiple consequents) and the 3-theory of dependent type 2-theories. Linearity as such isn’t the problem; the 3-theory of dependent type 2-theories is sufficiently general to include at least the straightforward dependent linear type theories where linear types depend on nonlinear types, and may also include at least some versions where linear types depend on each other. It’s the multiple-consequent aspect that it doesn’t handle. But yes, if trying to achieve such a unification one would have to also consider whether to try to model both prop and polycategorical composition or to choose only one.

Posted by: Mike Shulman on November 9, 2019 11:59 AM | Permalink | Reply to this

### Re: What Is an n-Theory?

A test for the numbering system:

• Philip Saville, Cartesian closed bicategories: type theory and coherence, (arXiv:2007.00624)

One might think the type theory here is a 3-theory since it concerns kinds of 2-category.

But there are traps, such as you mention here

Riehl-Shulman directed type theory is implicitly (sort of) about an $(\infty,2)$-category, but explicitly it is only about another particular kind of 1-categorical presentation, so it is again simply a 2-theory.

So is Saville’s type theory explicitly a 2-categorical presentation?

Posted by: David Corfield on July 2, 2020 1:51 PM | Permalink | Reply to this

### Re: What Is an n-Theory?

From a quick glance, it looks like yes, this is what I would call a 3-theory, since he has explicit separate syntax for 0-cells, 1-cells, and 2-cells (“rewrites”).

By the way, since Riehl-Shulman directed type theory has “interval variables” like cubical type theory, as I pointed out up here for the latter one might argue (although I’m not sure whether or not I would) that it is actually an $\infty$-theory.

Posted by: Mike Shulman on July 2, 2020 5:29 PM | Permalink | Reply to this

### Re: What Is an n-Theory?

And asking that such a function is a monoid homomorphism says precisely that this subset is a submonoid, i.e. contains the identity element $e\in M$ and is closed under the multiplication.

This is false. A function from M to $\Omega$ being a monoid homomorphism also requires that the product of an element not sent to “true” and another element (in either order) not be sent to “true”, which fails for eg the function from N to $\Omega$ sending even numbers to “true” and odd numbers to “false”, which indicates a submonoid but is not a monoid homomorphism since $f(2)=true\ne f(1)\wedge f(1)=false$.

Posted by: CatIsFluffy on March 10, 2023 7:22 AM | Permalink | Reply to this

### Re: What Is an n-Theory?

Presumably Mike’s version of submonoid does not count the even numbers as a submonoid of $\mathbb{N}$ since it doesn’t contain $1$.

This is standard, cf. the nLab definition of submonoid, and wikipedia’s definition.

Posted by: David Corfield on March 10, 2023 1:23 PM | Permalink | Reply to this

### Re: What Is an n-Theory?

I meant (N,+,0), not (N,*,1). For the latter, you get a similar problem with the submonoid consisting of the powers of 4.

Posted by: CatIsFluffy on March 10, 2023 9:32 PM | Permalink | Reply to this

### Re: What Is an n-Theory?

Right. So the only submonoids of $(N, +, 0)$ are $N$ itself and the singleton.

So is there a name for the kind of submonoid that’s wanted here?

Posted by: David Corfield on March 12, 2023 9:55 AM | Permalink | Reply to this

### Re: What Is an n-Theory?

You’re right. These are the submonoids $N$ of $M$ with the extra property that if $x y\in N$, then both $x\in N$ and $y\in N$. I don’t know a name for submonoids with this property, but there are at least some examples. For instance, if $M$ is the multiplicative monoid of a ring, then we can take $N$ to be the complement of any prime ideal in that ring.

Posted by: Mike Shulman on March 14, 2023 6:57 PM | Permalink | Reply to this

Post a New Comment