Two-dimensional Kripke Semantics I: Presheaves
Abstract
The study of modal logic has witnessed tremendous development following the introduction of Kripke semantics. However, recent developments in programming languages and type theory have led to a second way of studying modalities, namely through their categorical semantics. We show how the two correspond.
1 Introduction
The development of modal logic has undergone many phases [23, 15, 46, 98]. It is widely accepted that one of the most important developments was the relational semantics of Kripke [68, 69, 70] [15, �1] [46, �4.8]. Kripke semantics has proven time and again that it is intuitive and technically malleable, thereby exerting sustained influence over Computer Science.
However, over the last 30 years another way of studying modalities has evolved: looking at modal logic through the prism of the Curry-Howard-Lambek correspondence [72, 93, 99] yields new computational intuitions, often with surprising applications in both programming languages and formal proof. The tools of the trade here are type theory and category theory.
Up to now these two ways of looking at modalities have been discussed in isolation. The purpose of this paper is to establish a connection: I will show that the Kripke and categorical semantics of modal logic are part of a duality. It is well-known that dualities between Kripke and algebraic semantics exist: the Jónsson-Tarski duality is one of the cornerstones of classical modal logic [15, §5]. The main contribution of this paper is to show that such dualities can be elevated to the level of proofs. The punchline is that a profunctor , considered as a proof-relevant relation on a category , uniquely corresponds to a categorical model of modal logic on the category of presheaves on .
There are two obstacles to overcome to get to that result. The first is that we must work over an intuitionistic substrate: most research on types and categories is forced to do so, for unavoidable reasons [72, §8]. We must therefore first develop a duality for intuitionistic modal logic. However, there is no consensus on what intuitionistic modal logic is! The problem is particularly acute in the presence of [27]. I will avoid this problem by making canonical choices at each step. First, I will formulate a Kripke semantics based on bimodules, i.e. relations that are canonically compatible with a poset. Then, I will show how Kan extension uniquely determines two adjoint modalities, and , from any bimodule. The fact these arise automatically is evidence that they are the canonical choice of intuitionistic modalities.
The second obstacle stems from considering proofs. The jump from algebraic to categorical semantics involves adding an extra ‘dimension’ of proofs. Consequently, in order to re-establish a duality, an additional dimension must be added to Kripke semantics as well. I call the result a two-dimensional Kripke semantics. Category theorists will find it anticlimactic: it amounts to the folklore observation that a proof-relevant Kripke semantics is essentially a semantics in a presheaf category.
Indeed, a sizeable proportion of this paper consists of folklore results that are well-known to experts. However, many of them are drawn from related but distinct areas: logic, order theory, category theory, and topos theory. As a result, it does not appear that all of them are well-known by a single expert. Thus, the synthesis presented here appears to be new.
The results I present in this paper show that there are deep connections between modal logic and presheaf categories. This is important, as the latter are ubiquitous in logic and related fields: presheaf models are used in fields as disparate as categorical homotopy theory [87, 24], type theory [57], concurrency [65, 21, 22], memory allocation [81, 82], synthetic guarded domain theory [14], second-order syntax and algebraic theories [34, 53, 35, 36, 37], higher-order abstract syntax [58], and so on. As a result, the connections presented here may enable synthetic reasoning via modalities in a variety of logical settings.
In § 2 I recall the Kripke and algebraic semantics of intuitionistic logic, and outline the duality between Kripke semantics and certain complete Heyting algebras, the prime algebraic lattices. Then I extend this duality to intuitionistic modal logic in § 3 by showing how a relation that is compatible with the intuitionistic order—a bimodule—gives rise to modalities through Kan extension. In § 4 I add proofs to intuitionistic logic, and elevate the duality to one between ‘two-dimensional frames’ and presheaf categories. I then repeat this exercise for intuitionistic modal logic in § 5 by promoting bimodules to profunctors on the relational side, and adding an adjunction on the categorical side.
For general background in orders please refer to the book by Davey and Priestley [28]. Given a poset let the opposite poset be given by reversing the partial order; that is, iff . A lattice has all finite meets and joins. A complete lattice has arbitrary ones. A complete lattice is infinitely distributive just if the law holds. Such lattices are variously called frames, locales, or complete Heyting algebras [61, 76, 84].
2 Intuitionistic Logic I
There are many types of semantics for intuitionistic logic, including Kripke, Beth, topological, and algebraic semantics. Bezhanishvili and Holliday [11] argue that these form a strict hierarchy, with Kripke being the least general, and algebraic the most general. I will briefly review the elements of these extreme points of the spectrum.
The Kripke semantics of intuitionistic logic are given by Kripke frames, i.e. partially-ordered sets [23, §2.2]. is referred to as the set of possible worlds, and as the information order. A world is a ‘state of knowledge,’ and means that moving from to potentially entails an increase in the amount of information.
Let be the set of upper sets of , i.e. the subsets such that and implies . A Kripke model consists of a Kripke frame as well as a function . The valuation assigns to each propositional variable an upper set , which is the set of worlds in which is true. The idea is that, once a proposition becomes true, it must remain true as information increases.
We are now able to inductively define a relation with the meaning that is true in world of model . The only interesting clause is that for implication:
This definition is famously monotonic: if and then . Kripke semantics is sound and complete for intuitionistic logic [23, 15].
The algebraic semantics of intuitionistic logic consist of Heyting algebras. These are lattices such that every map has a right adjoint, i.e. for there is an element such that iff . Such lattices are always distributive. Assuming that one has an interpretation of each proposition , each formula of intuitionistic logic is inductively mapped to an element using the corresponding algebraic structure. I will not expound further on Heyting algebras; see [23, §7.3] [16, §1.1] [76, §I.8]. But I note that they are sound and complete for intuitionistic logic.
2.1 Prime algebraic lattices
Let be any Kripke frame, and let . Consider the poset of monotonic functions from to , ordered pointwise. This poset has a number of curious properties.
First, the monotonicity of means that if and , then . Hence, the subset of is an upper set. Conversely, every upper set gives rise to a monotonic by setting if , and otherwise. Consequently, there is an order bijection
with the order on being inclusion. I will liberally treat upper sets and elements of as the same, but prefer the latter notation for reasons that will become clear later.
Second, given any , consider its principal upper set . This set consists of worlds with potentially more information than that found in world . A simple argument shows that iff .111This is an order-theoretic consequence of the Yoneda lemma. Thus, this gives an order embedding
which can be shown to preserve meets and exponentials.
Third, the poset is a complete lattice: arbitrary joins and meets are given pointwise. Viewing the elements of as upper sets, these joins and meets correspond to arbitrary unions and intersections of upper sets, which are also upper. Moreover, this lattice satisfies the infinite distributive law, so it is a complete Heyting algebra—synonymously a frame or locale [61, 84]. Given two upper sets their exponential is given by [29, §1.9]
Fourth, the principal upper sets are special, in that they are prime.222Such elements are variously called completely join-irreducible [86], supercompact [8] [84, §VII.8], completely (join-)prime [100], or simply join-prime [41, §1.3]. An element of a complete lattice is prime just if
This says that contains a tiny, indivisible fragment of information: as soon as it approximates a supremum, it must approximate something in the set that is being upper-bounded. The prime elements of are exactly the principal upper sets for some .
Fifth, the complete lattice is prime algebraic. This means that all its elements can be reconstructed by ‘multiplying’ or ‘sticking together’ prime elements. In symbols, a complete lattice is prime algebraic whenever for every element we have
Such lattices are variously called completely distributive, algebraic lattices [28, §10.29] or superalgebraic lattices [84, §VII.8]. In fact, it can be shown that any such lattice is essentially of the form , i.e. a lattice of upper sets; this was shown by Raney in the 1950s [86], and independently by Nielsen, Plotkin and Winskel in the 1980s [80]. See the paper by Winskel for the use of prime algebraic lattices in semantics [100].
Finally, the fact every element can be reconstructed as a supremum of primes means that it is possible to canonically extend any monotonic to a monotonic , as long as is a complete lattice. Diagrammatically, in the situation
| (1) |
there exists a unique which preserves all joins and satisfies . It is given by
is called the (left) Kan extension of along . As preserves all joins and is complete it has a right adjoint , by the adjoint functor theorem [28, §7.34] [61, §I.4.2]. For any complete lattice this situation amounts to a bijection
where CSLatt is the category of complete lattices and join-preserving maps.
Suppose then that we have a Kripke model . The construction given above induces a Heyting algebra . Defining we obtain an algebraic model of intuitionistic logic, which interprets every formula as an upper set . This is the upper set of worlds in which a formula is true [23, Theorem 7.20]:
Theorem 2.1.
if and only if .
Thus, every Kripke semantics corresponds to a prime algebraic lattice.
Remark 2.2.
This shows that a Kripke semantics is a particular kind of algebraic semantics. Thus, we can deduce the completeness of the latter from the completeness of the former: if a formula is valid in all Heyting algebras, it must be valid in all prime algebraic lattices, and hence valid in all Kripke semantics. If the Kripke semantics is complete, then the formula must be provable. Therefore, the algebraic semantics is then complete as well.
The opposite direction—viz. proving the completeness of Kripke semantics from completeness of the algebraic semantics—cannot be shown constructively. The reason is that it requires the construction of prime filters, which is a weak form of choice. I will investigate the details of this mismatch in a sequel paper.
2.2 Morphisms
The simplest kind of morphism between Kripke frames is a monotonic map . Frames and monotonic maps form the category Pos of posets. Given a monotonic we can define a monotonic by taking to . Viewing the elements of as upper sets, maps the upper set to the set , which is upper by the monotonicity of . preserves arbitrary joins and meets. It is thus the morphism part of a functor to the category PrAlgLatt of prime algebraic lattices and complete lattice homomorphisms.
Moreover, the functor is an equivalence! By the adjoint functor theorem any complete lattice homomorphism has a left and right adjoint:
| (2) |
Given a prime algebraic lattice , let be the sub-poset of prime elements. It can be shown that the left adjoint maps primes to primes [41, Lemma 1.23]. We can thus restrict it to a function . This defines a functor with the property that . All in all, this amounts to a duality
| (3) |
However, monotonic maps are not particularly well-behaved from the perspective of logic, as they do not preserve nor reflect ‘local’ truth. This is the privilege of open maps.
Definition 2.3.
Let map the unique point of to . A monotonic map of Kripke frames is open just when it has the right lifting property with respect to , i.e. when every commuting diagram of the form
in Pos has a diagonal filler (dashed) that makes it commute.
In other words, is open if whenever there exists a with and .333Such morphisms are often called p-morphisms [23, §2.3] or bounded morphisms [15, §2.1]. According to Goldblatt [46] open maps were introduced by de Jongh and Troelstra [29] in intuitionistic logic, and by Segerberg [91] in modal logic. More rarely they are called functional simulations, and led us to bisimulations [90, §3.2]. The name is chosen because such maps are open with respect to the Alexandrov topology on a poset, whose open sets are the upper sets [61, §1.8]. Open maps send upper sets to upper sets [23, Prop. 2.13]. Thus
Lemma 2.4.
Let and be Kripke models, and be open. Suppose , i.e. iff . Then iff .
Write to mean that for any valuation and . Then
Lemma 2.5.
If is open and surjective, then implies .
Recall now the induced map for a monotonic . The following lemma allows us to characterise the openness and surjectivity of in terms of .
Lemma 2.6.
-
1.
is open iff preserves exponentials.
-
2.
is surjective iff is injective.
Consequently, the duality (3) may be restricted to two wide subcategories:
| (4) |
The morphisms of the categories to the left of are open (resp. and surjective) maps, and the morphisms of the categories to its right are complete Heyting homomorphisms, i.e. complete lattice homomorphisms that preserve exponentials (resp. and are injective).
Finally, let us consider the classical case—as a sanity-check. This amounts to restricting Pos to its subcategory of discrete orders, i.e. . In this case every map is open. The corresponding restriction on the other side is to the category CABA of complete atomic Boolean algebras, yielding the usual Tarski duality [67].
2.3 Related work
The origins of the construction of a Heyting algebra from a Kripke frame seems to be lost in the mists of time. The earliest occurrence I have located is in the book by Fitting [39, §1.6], where it is attributed to an exercise in the book by Beth [10].
The duality (3) appears to be folklore—folklore enough to be included as an exercise in new textbooks [41, Ex. 1.3.10]; see also Erné [31]. However, I have not been able to find any mention of the dualities of (4) in the literature.
Both the dualities (3) and (4) involve just prime algebraic lattices, which is a far cry from encompassing all Heyting algebras. It is possible to do so, by enlarging the category Pos to a class of ordered topological spaces called descriptive frames [23, §8.4]. The resulting duality is called Esakia duality [32] [41, §4.6] [12, §2.3].
A survey on dualities for classical modal logic is given by Kishida [67].
3 Modal Logic I
The task now is to extend the results of § 2 to intuitionistic modal logic.
There is disagreement on what a minimal intuitionistic modal logic is. This arises no matter the methodology we choose—be it relational, algebraic, or proof-theoretic. The situation becomes even more complex if we include a diamond modality (): see Das and Marin [27] and Wolter and Zakharyaschev [103] for a discussion.
In this paper I will adopt the intuitionistic propositional logic with Galois connections of Dzik, Järvinen, and Kondo [30], for reasons that will become clear in a moment. This extends intuitionistic logic with modalities and , and the two inference rules
| and |
These rules correspond to a Galois connection [28, §7.23], i.e. an adjunction between posets. They imply the derivability of the following rules, amongst others [30, Prop. 2.1].
The notation of the ‘black diamond’ modality may appear unusual. However, I will argue that this logic is, in a way, the canonical intuitionistic modal logic.
The Kripke semantics of classical modal logic is given by a modal frame , which consists of a set and an arbitrary accessibility relation [15, §1]. If the same set of worlds is already part of an intuitionistic Kripke frame we must take care to ensure that and are compatible. There are many compatibility conditions that one can consider [85] [92, §3.3]. However, I will take a hint from the category theory literature, and seek a canonical definition of what it means for a relation to be compatible with a poset.
Recall that relations can be presented as functions which map a pair of worlds to iff . I will ask that is such function, but with a twist:
Definition 3.1.
A bimodule is a monotonic map .
Thus, a relation is a bimodule just if implies . This means that can absorb changes in information on either side: contravariantly on the first component, and covariantly on the second. This is a standard, minimal way to define what it means to be ‘a relation in Pos.’ It is strongly reminiscent of bimodules in abstract algebra.
We can then define a modal Kripke frame to be a Kripke frame equipped with a bimodule . A modal Kripke model adds to this a function . We extend to modal formulae:
There are a number of things to note about this definition. First, there is a deep duality between the clauses: not only do we exchange for , but we also flip the variance of . As a result, uses the relation in the opposite variance to the more traditional modality (hence the change in notation). Second, the clause for the modality is the traditional one; some streams of work on intuitionistic modal logic adopt a slightly different one [85, 92], which is equivalent to this in the presence of the bimodule condition. Finally, this definition is monotonic: the bimodule conditions on suffice to show that if and then . Dzik et al. [30, §5] prove that this semantics is sound and complete.
The algebraic semantics of this logic is given by a Heyting algebra equipped with two monotonic maps which form an adjunction , i.e. a Galois connection. Dzik et al. [30, §4] prove that this semantics is also sound and complete.
We are now in a position to relate the Kripke and algebraic semantics of this intuitionistic modal logic. Let be a modal Kripke frame, and consider the map obtained by cartesian closure of Pos. This map takes to the upper set of worlds accessible from . Putting in (1) we obtain through Kan extension the diagram
| (5) |
where we write for and for . It can be shown that these maps are given by
Thus, any bimodule defines an adjunction on . Correspondingly, any adjunction on yields a monotonic map , which uniquely corresponds to a bimodule by the cartesian closure of Pos.
Thus, starting from a bimodule, i.e. a relation that is compatible with the information order, we have canonically obtained a model of intuitionistic modal logic on through Kan extension: is a complete Heyting algebra, and we define and . We immediately obtain a modal analogue to Theorem 2.1:
Theorem 3.2.
For any modal formula , if and only if .
3.1 Morphisms
We define a category Bimod with bimodules as objects. A bimodule morphism from to is a pair of monotonic maps and such that . Stated in terms of relations, it must be that implies .
We define the subcategory EBimod to consist of (endo)bimodules and pairs of maps . Thus, objects are bimodules on a single poset , and morphisms are monotonic maps that preserve the relation, i.e. implies . In other words, the objects of EBimod are modal Kripke frames, and the morphisms are monotonic, relation-preserving maps.
Recall the adjunctions and modalities induced by a monotonic :
| (6) |
Lemma 3.3.
is a morphism of bimodules iff .
This constitutes a duality
| (7) |
where PrAlgLattO is the category with objects , where is a prime algebraic lattice and is an operator that preserves all meets. By the adjoint functor theorem, such operators always have a left adjoint . Thus, this category contains algebraic models of intuitionistic modal logic (but not all of them). By the preceding section each such adjunction corresponds uniquely to a bimodule. The morphisms of PrAlgLattO are complete lattice homomorphisms such that . By the preceding lemma they correspond precisely to morphisms of bimodules.
However, as with monotone maps, morphisms of bimodules do not preserve local truth; for that we need a notion of modally open maps.
Definition 3.4.
Let and be modal Kripke frames. A bimodule morphism is modally open just if whenever then there exists a with and .
This is similar to Definition 2.3, but ever so slightly weaker: instead of requiring , it requires that the information in can be increased to . Like Definition 2.3, it can also be written homotopy-theoretically, but that requires some ideas from double categories that are beyond the scope of this paper. We have the analogous result about preservation of truth:
Lemma 3.5.
Let and be modal Kripke models, be open and modally open, and . Then iff .
Lemma 3.6.
Let be open, modally open, and surjective. If then .
The following result relates the modal openness of to .
Lemma 3.7.
is modally open iff iff .
Thus, the duality (7) may be restricted to dualities between wide subcategories:
| (8) |
The morphisms to the left of are open and modally open (resp. and surjective); and the to the right of it preserve exponentials and commute with operators (resp. and are injective).
Let us consider the restriction of this duality to the classical setting—as a sanity check. A bimodule on a discrete poset is just a relation on a set. The corresponding restriction on the right is to CABAs with operators, and complete homomorphisms which commute with operators. We thus obtain the Thomason duality between Kripke frames and modally open maps on the left, and CABAs with operators to the right [96, 67].
3.2 Related work
Many works have presented a Kripke semantics for intuitionistic modal logic. All such semantics assume two accessibility relations: a preorder for the intuitionistic dimension, and a second relation for the modal dimension. What varies is their compatibility conditions.
The first work to present such a semantics appears to be that of Fischer Servi [38]. One of the required compatibility conditions is . This is weaker than having a bimodule, but sufficient to prove soundness.
The first work to recognise the importance of bimodules was Sotirov’s 1979 thesis. His results are summarised in a conference abstract [94, §4]: they include the completeness of a minimal intuitionistic modal logic with a , the K axiom, and the necessitation rule. Božić and Došen [19] repeat the study for the same logic, but for a semantics based on the Fischer Servi compatibility conditions. However, they note that their completeness proof actually constructs half a bimodule (a ‘condensed’ relation). They also point out that bimodules, which they call ‘strictly condensed’ relations, are sound and complete for their logic. Wolter and Zakharyaschev [101, §2] argue that bimodule and Fischer Servi semantics are equi-expressive.
Plotkin and Stirling [85] attempt to systematise the Kripke semantics of intuitionistic modal logic. Their frame conditions allow ‘transporting a modal relation upwards’ along any potential increases of information on either side. This paper and all its descendants—notably the thesis of Simpson [92, §3.3]—adopt a different satisfaction clause for which uses both and . In the presence of the bimodule conditions this satisfaction clause is equivalent to the classical one, which I use here.
The bimodule condition and the complex algebra construction (or fragments thereof) have made scattered appearances in the literature: in the early work of Sotirov [94] and Božić and Došen [19]; in the work of Wolter and Zakharyaschev [102, 101, 103], Hasimoto [54, §4], and Orłowska and Rewitzky [83]; and of course in Dzik et al. [30, §7]. With the exception of the last one, none of these references discuss the modality. Moreover, in none of these references are the categorical aspects of this construction discussed.
As mentioned before, dualities between frames and algebras have played a significant role in modal logic. Thomason [96] and Goldblatt [45] also considered morphisms of frames, respectively obtaining Thomason duality and (categorical) Jónsson-Tarski duality between descriptive frames and Boolean Algebras with Operators (BAOs) [46, §6.5]. Kishida [67] surveys a number of dualities for classical modal logic.
The duality (7) is stated by Gehrke [40, Thm. 2.5] who attributes it to Jónnson [64], even though no such theorem appears in that paper.
The dualities of (8) are the direct intuitionistic analogues to that of Thomason. I have not been able to find them anywhere in the literature.
According to the extensive survey of Menni and Smith [78], the idea that the commonly-used modalities and are often part of adjunctions and is implicitly present throughout the development of modal logic. However, these were not made explicit in a logic until the 2010s, when they appeared in the work of Dzik et al. [30] and Sadrzadeh and Dyckhoff [89]. The same perspective plays a central rôle in the exposition of Kishida [67].
4 Intuitionistic Logic II
In the rest of this paper we will categorify [7] the notion of Kripke semantics. The main idea is to replace posets by categories, so that the order is replaced by a morphism . As there might be multiple morphisms , this allows the recording of not just the fact may signify more information than , but also the manner in which it does so. The reflexivity and transitivity of the poset are then replaced by the identity and composition laws of the category. This adds a dimension of proof-relevance to Kripke semantics.
A corresponding change in our algebraic viewpoint will be that of replacing the set of truth values with the category . This is a classic Lawverean move [73]. Notice that this is lopsided, as is usual in intuitionistic logic: while the falsity is only represented by one value, viz. the empty set, the truth can be represented by any non-empty set . The elements of can be thought of as a proofs of a true statement.
Let us then trade the frame for an arbitrary category . It remains to define what it means to have a proof that the formula holds at a world . We denote the set of all such proofs by . Assuming we are given a set for each proposition and world , here is a first attempt:
where for a family of sets we let
This closely follows the usual Kripke semantics, but adds proofs. For example, a proof in is a pair of a proof and a proof . Similarly, a proof is a function which maps a proof of ‘increase in information’ to a function . In turn, this function maps proofs in to proofs in .
To show that this definition is monotonic we have to demonstrate it on proofs: given a proof and a morphism we have to define a proof . Assuming that we are given this operation for propositions, we can extend it by induction; e.g.
Moreover, this definition is compatible with , in the sense that and . We thus obtain a (covariant) presheaf for each formula .
It is well-known that the proofs of intuitionistic logic form a bicartesian closed category (biCCC), i.e. a category with finite (co)products and exponentials [71]. A biCCC can be seen as a categorification of a Heyting algebra: formulae are objects of the category, and proofs are morphisms. We will not expound on this further; see [72, 26, 4].
It should therefore be the case that the semantics described above form a biCCC. Indeed, it is a well-known fact of topos theory that the category of presheaves is a biCCC. In fact, the construction of exponentials [76, §I.6] reveals that our definition above is deficient: we should restrict to contain only those functions that satisfy a naturality condition, i.e. those which for any , , and satisfy
From this point onwards I will identify two-dimensional Kripke semantics with categorical semantics in a category of presheaves .
4.1 Presheaf categories
The category of covariant presheaves is eerily similar to prime algebraic lattices. In a sense they are just the same; but, having traded for , they have become proof-relevant.
First, letting , an element is a proof that holds at a ‘world’ . A morphism of then leads to a proof that holds at . Thus, the presheaf is very much like an upper set.
Second, the representable presheaves are the proof-relevant analogues of the principal upper set. By the Yoneda lemma they constitute an embedding
which moreover preserves limits and exponentials [4].
Third, the category is both complete and cocomplete, with limits and colimits computed pointwise [76, §I]. It is also ‘distributive’ in an appropriate sense [3, §3.3], which makes it into a Grothendieck topos. It is thus a cartesian closed category, with exponential
which is essentially the two-dimensional semantics of implication I gave above.
Fourth, the representables are special, in that they are tiny [104].
Definition 4.1.
An object is tiny just if preserves colimits.444In the literature this property is often referred to as external tininess (cf. internal tininess).
Tininess is a proof-relevant version of primality: it implies that for any there exists an such that is equal to the composition of a morphism with the injection . By the Yoneda lemma it follows that all representables are tiny, as they satisfy the above definition for and .
Fifth, the so-called co-Yoneda lemma [75, §III.7] shows that every is a colimit of representables. This means that it can be reconstructed by sticking together tiny elements:
Like with prime algebraic lattices, there is a converse to this result: every category which is generated by sticking together tiny elements is in fact a presheaf category:
Theorem 4.2 (Bunge [20]).
A category which is cocomplete and strongly generated by a small set of tiny objects is equivalent to for some small category .
A textbook presentation of this result can be found in the book by Kelly [66, §5.5].
Finally, the fact every element can be reconstructed as a colimit of representables means that it is possible to uniquely extend any functor to a cocontinuous functor , as long as is cocomplete. Diagrammatically, in the situation
| (9) |
there exists an essentially unique cocontinuous with . It is given by
is called the left Kan extension of along . It has a right adjoint which is explicitly given by . This amounts to an isomorphism
where is the category of categories, and Cocont is the category of cocomplete categories and cocontinuous functors: see [4, Prop. 9.16] [88, Cor. 6.2.6, Rem. 6.5.9] and [75, § X.3, Cor. 2] [66, Th. 4.51].
| poset | category |
| monotonic map | functor |
| upper sets | presheaves |
| principal upper set | representable presheaf |
| prime element | tiny object |
| prime algebraic lattice | presheaf category |
| bimodule | profunctor |
All in all, presheaf categories are the categorification of prime algebraic lattices.
4.2 Cauchy-complete and spacelike categories
Replacing posets with categories does not come for free: the extra dimension of morphisms leads to situations that have no analogues in poset. Some of these are problematic when thinking of as a two-dimensional Kripke frame. Perhaps the most bizarre is the presence of idempotents, i.e. morphisms with the property that . Such morphisms represent a non-trivial increase in information which confusingly leaves us in the same world.
The presence of idempotents causes issues. For example, recall that, in prime algebraic lattices, primes and principal upper sets coincide. The astute reader will have noticed we did not claim the analogous result in presheaf categories: tiny objects are not necessarily representable in . For that, we need to be Cauchy-complete [18, 17].
Definition 4.3.
A category is Cauchy-complete just if every idempotent splits, i.e. if every idempotent is equal to for a section-retraction pair and .
Note that every complete category is Cauchy-complete, including and .
This leads us to another troublesome situation, namely that of having section-retraction pairs, i.e. and with . In this case and contain no more information than each other, but are not isomorphic. We may ask that this does not arise.
Definition 4.4.
A category satisfies the Hemelaer condition [55, Prop. 5.8] just if every section-retraction pair is an isomorphism.
Combining these two conditions is equivalent to the following definition.
Definition 4.5.
A category is spacelike if every idempotent is an identity.
Lawvere has identified this condition as having particular importance in recognising petit toposes [77]. We will not use it much, as it restricts the dualities we wish to develop.
In the rest of this paper we will assume that our base categories are Cauchy-complete, so that tiny objects coincide with representables.
4.3 Morphisms
The simplest kind of morphism between categories is a functor. Given a we can define a functor that takes to . This functor has left and right adjoints, which are given by Kan extension [62, A4.1.4]:
| (10) |
Therefore preserves all limits and colimits, i.e. it is (co)continuous. In short, the presheaf construction gives a functor , where is the category of small Cauchy-complete categories and functors; and PshCat is the category of presheaf categories (over Cauchy-complete base categories) and (co)continuous functors.
Moreover, this functor is an equivalence. Given a presheaf category we can obtain its base as the subcategory of tiny objects [62, A1.1.10]. But how can we extract from any (co)continuous functor ? First, as presheaf categories are locally presentable, the adjoint functor theorem implies that has left and right adjoints, as in (10) [1, §1.66]. This gives what topos theorists call an essential geometric morphism. Johnstone [62, §A4.1.5] shows that every such morphism is induced by a , as preserves representables (when is Cauchy-complete). We thus obtain a duality
| (11) |
As with posets, functors here fail to preserve truth; for that we need a notion of openness.
Definition 4.6.
is open just if preserves exponentials.
Lemma 4.7.
If is open then there is a natural isomorphism .
Definition 4.6 is somewhat underwhelming, as it does not give explicit conditions that one can check—unlike Definition 2.3. However, obtaining such a description appears difficult.
Some information may be gleaned by considering as a geometric morphism. Such a morphism is open [60] [63, C3.1] just if both the canonical maps and are monic. Johnstone [63, C3.1] proves that is open iff for any in there exists an in and a section-retraction pair and with . This superficially seems like a categorification of Definition 2.3. However, it only guarantees that is sub-cartesian-closed, whereas we need an isomorphism for Lemma 4.7 to hold.
A stronger condition is to ask that be locally connected, i.e. that commute with dependent products [63, C3.3]. All such morphisms are open geometric morphisms. This is stronger than what we need, but sufficient conditions on can be given [63, C3.3.8].
Finally, an even stronger condition is to ask that be atomic, i.e. that is a logical functor. This means it preserves exponentials and the subobject classifier [63, A2.1, C3.5]. All atomic geometric morphisms are locally connected. This is again stronger than what we need, and a characterisation in terms of is elusive: see MathOverflow [95].
It is easier to characterise when is a surjective geometric morphism, i.e. when is faithful [62, A2.4.6]. This happens exactly when is retractionally surjective, i.e. whenever every is the retract of for some [62, A2.4.7]. If satisfies the Hemelaer condition this reduces to being essentially surjective.
Write to mean that is non-empty for any and any interpretation of .
Lemma 4.8.
Let be open and retractionally surjective. If then .
We may thus restrict the duality (11) to dualities
| (12) |
In the first instance the category to the left of is that of small Cauchy-complete categories and open functors; and to the right of it is presheaf categories and (co)complete, cartesian closed functors. In the second instance the category to the left of is that of small Cauchy-complete categories and open, retractionally surjective functors; and to the right of it is presheaf categories and (co)complete, faithful, cartesian closed functors.
5 Modal Logic II
To make a two-dimensional Kripke semantics for modal logic we have to categorify relations. We took the first step by considering bimodules, i.e. information-order-respecting relations. The second step can be taken by replacing with ; this leads us to the notion of a relation between categories, also known as a profunctor or distributor [9] [17, §7].
Definition 5.1.
A profunctor is a functor .
To formulate a two-dimensional Kripke semantics for modal logic we replace modal Kripke frames with a small Cauchy-complete category with an (endo)profunctor . To obtain the modalities we can now play the same trick: putting into (9) we canonically obtain the following diagram by Kan extension:
| (13) |
Conversely, any adjunction on corresponds to the (endo)profunctor on given by .
We may then define and . It is worth unfolding what a proof of at a world is to obtain an explicit description:
| (14) |
Thus, a proof that holds at is a natural transformation . This has the expected shape of Kripke semantics for : for each and proof that is accessible from it gives us a proof that holds at .
It is a little harder to see what a proof of at a world is. It becomes more perspicuous if we use the coend formula for the left Kan extension [74, §2.3]:
| (15) |
Hence, a proof that holds at consists of a world , a proof that , and a proof that holds at —which is exactly what one would expect. The difference is that the coend quotients some of these pairs according to the action of . See Mac Lane and Moerdijk [76, §VII.2] for a textbook exposition on why this is a tensor product of and .
How well does this fit the categorical semantics of modal logic? As with intuitionistic modal logic, there is also a number of proposals of what that might be. A fairly recent idea is to define it as the semantics of a Fitch-style calculus, as studied by Clouston [25]. This is exactly a bicartesian closed category equipped with an adjunction:
| (16) |
The left adjoint is often written as lock. It does not commonly appear as a modality, but as an operator on contexts that corresponds to ‘opening a box’ in Fitch-style natural deduction [59, §5.4]. The modality is a right adjoint, so that it automatically preserves all limits, including products. This idea has proven remarkably robust: variations on it have worked well for modal dependent type theories [13, 51, 52, 50, 49]. The fact that an adjunction on a presheaf category corresponds precisely to a two-dimensional Kripke semantics is further evidence that this is the correct notion of categorical model of modal logic.
Finally, note that (14) and (15) look suspiciously similar to the modal structure of Normalization-by-Evaluation models for modal type theories. This is explicitly visible in the paper by Valliappan et al. [97, §2], and also implicitly present in the paper by Gratzer [48].
5.1 Morphisms
Define the category Prof to have as objects profunctors. A morphism from to consists of functors and , and a natural transformation . The subcategory EProf consists of endoprofunctors , and triples of the form . I will synecdochically refer to as a morphism of EProf. Thus, objects are two-dimensional Kripke frames, and morphisms are functors that proof-relevantly preserve the relation.
Lemma 5.2.
Morphisms of endoprofunctors are in bijection with natural transformations .
Proof 5.3.
Unfolding the definitions, . As this is exactly a transformation . By the Yoneda lemma, any such transformation arises by precomposition with a unique transformation . By again, this uniquely corresponds to a transformation .
We thus obtain a duality
| (17) |
where PshCatO is the category of presheaf categories equipped with a continuous . Note that, as presheaf categories are locally finitely presentable, always has a left adjoint . Thus, the objects are categorical models of modal logic. Morphisms are pairs of a (co)continuous and a natural transformation .
As before, open functors do not preserve truth; for that we need a notion of modal openness. Let . As pointed out in the proof of Lemma 5.2 such an uniquely corresponds to a transformation . Its components
map and to . We can then say that
Definition 5.4.
is modally open just if is an isomorphism.
This asks that for every proof we should be able to find an object , a proof , and a morphism , so that . This is clearly a categorification of Definition 3.4, and leads to the following lemma:
Lemma 5.5.
is modally open iff the corresponding is an isomorphism.
Proof 5.6.
The proof of Lemma 5.2 precomposes with to get . Thus is iso iff is.
Thus, the duality (17) may be restricted to dualities between the wide subcategories
| (18) |
The morphisms to the left of are modally open, open maps (resp. and retractionally surjective); and the morphisms to the right of are where is cartesian closed (resp. and faithful) and is a natural isomorphism.
6 Other related work
Perhaps the work most closely related to this paper is that on Kripke-style lambda models by Mitchell and Moggi [79]. These amount to elaborating the first-order definitions of applicative structure and -model in the internal language of a presheaf category, with the base category being a partial order. In practice this means that the interpretation of function types is only a subfunctor of the exponential of presheaves [79, §8]. However, Mitchell and Moggi prove that these models are sound and complete for the fragment, even in the presence of empty types. They also use some general theorems about open geometric morphisms to prove that any cartesian closed category can be presented as such a model.
Another piece of work that bears kinship with the present one is Hermida’s fibrational account of relational modalities [56]. Hermida shows that both the relational modalities and can be obtained canonically as extensions of predicate logic to relations, with the modalities arising as compositions of adjoints. The black diamond makes a brief cameo as the induced left adjoint to , as does the dual black box [56, §3.3]. While the decompositions obtained by Hermida seem more refined than the results here, Kan extension does not make an explicit appearance. As such, the relationship to the present work is yet to be determined.
Awodey and Rabe [6] give a Kripke semantics for extensional Martin-Löf type theory (MLTT), in which contexts are posets and types are presheaves over them. They use topos-theoretic machinery to prove that every locally cartesian closed category can be embedded in a presheaf category over a poset; this result seems similar to one of Mitchell and Moggi, but the proof appears entirely different. As a consequence, they show that presheaf categories over posets form a complete class of models for extensional MLTT, in fact a subclass of locally cartesian closed categories.
Alechina et al. [2] present dualities between Kripke and algebraic semantics for constructive S4 and propositional lax logic. Their interpretation of follows that of Plotkin, Stirling and Simpson [85, 92].
Ghilardi and Meloni [42] explore a presheaf-like interpretation of (predicate) modal logic, which is similar to ours, albeit non-proof-relevant. They work over the identity profunctor . They are hence forced to weaken the definition of presheaf. See also [43, 44].
Awodey, Kishida and Kotzsch [5] give a topos-theoretic semantics for a higher-order version of intuitionistic S4 modal logic. They also briefly survey much previous work on presheaf-based and topos-theoretic semantics for first-order modal logic. Their work is not proof-relevant.
Finally, there is clear methodological similarity between the results obtained here and the results of Winskel and collaborators on open maps and bisimulation [65, 22]. One central difference is that Winskel et al. are mainly concerned with open maps between presheaves themselves, whereas I only consider open maps between (two-dimensional) frames.
Acknowledgements
I have benefitted significantly from conversations with Dan Licata, Nachiappan Valliappan, Fabian Ruch, Amar Hadzihasanovic, Kohei Kishida, Sean Moss, Sam Staton, Daniel Gratzer, Lars Birkedal, Jonathan Sterling, Philip Saville, and Gordon Plotkin.
This work was supported by the UKRI Engineering and Physical Sciences Research Council grants EP/Y000242/1, EP/Y033418/1, the UKRI International Science Partnerships Fund (ISPF), and a Royal Society Research Grant.
References
- [1] Jiří Adámek and Jiří Rosický. Locally Presentable and Accessible Categories. Cambridge University Press, 1994. doi:10.1017/CBO9780511600579.
- [2] Natasha Alechina, Michael Mendler, Valeria de Paiva, and Eike Ritter. Categorical and Kripke semantics for constructive S4 modal logic. Computer Science Logic, pages 292–307, 2001. doi:10.1007/3-540-44802-0_21.
- [3] Mathieu Anel and André Joyal. Topo-logie. In Mathieu Anel and Gabriel Catren, editors, New Spaces in Mathematics, pages 155–257. Cambridge University Press, 2021. doi:10.1017/9781108854429.007.
- [4] Steve Awodey. Category Theory. Oxford Logic Guides. Oxford University Press, 2010.
- [5] Steve Awodey, Kohei Kishida, and Hans-Christoph Kotzsch. Topos Semantics for Higher-Order Modal Logic. Logique et Analyse, 228:591–636, 2014. doi:10.2143/LEA.228.0.3078176.
- [6] Steve Awodey and Florian Rabe. Kripke Semantics for Martin-Löf’s Extensional Type Theory. Logical Methods in Computer Science, 7(3):1–34, 2011. doi:10.2168/LMCS-7(3:18)2011.
- [7] John C. Baez and James Dolan. Categorification. In Ezra Getzler and Mikhail Kapranov, editors, Higher Category Theory, number 230 in Contemporary Mathematics. American Mathematical Society, 1998. eprint: math/9802029. doi:10.48550/arXiv.math/9802029.
- [8] B. Banaschewski and G. Bruns. The fundamental duality of partially ordered sets. Order, 5(1), 1988. doi:10.1007/BF00143898.
- [9] J. Bénabou. Distributors at work, 2000. Notes of Thomas Streicher based on lectures of Jean Bénabou. URL: https://www2.mathematik.tu-darmstadt.de/~streicher/FIBR/DiWo.pdf.
- [10] Evert Willem Beth. The Foundations of Mathematics. North-Holland Pub. Co., 1959.
- [11] Guram Bezhanishvili and Wesley H. Holliday. A semantic hierarchy for intuitionistic logic. Indagationes Mathematicae, 30(3):403–469, 2019. doi:10.1016/j.indag.2019.01.001.
- [12] Nick Bezhanishvili. Lattices of intermediate and cylindric modal logics. PhD thesis, University of Amsterdam, 2006. Report number DS-2006-02. URL: https://eprints.illc.uva.nl/id/eprint/2049/.
- [13] Lars Birkedal, Ranald Clouston, Bassel Mannaa, Rasmus Ejlers Møgelberg, Andrew M. Pitts, and Bas Spitters. Modal dependent type theory and dependent right adjoints. Mathematical Structures in Computer Science, 30(2):118–138, 2020. doi:10.1017/S0960129519000197.
- [14] Lars Birkedal, Rasmus Møgelberg, Jan Schwinghammer, and Kristian Støvring. First steps in synthetic guarded domain theory: step-indexing in the topos of trees. Logical Methods in Computer Science, 8(4), 2012. doi:10.2168/LMCS-8(4:1)2012.
- [15] Patrick Blackburn, Maarten de Rijke, and Yde Venema. Modal Logic. Cambridge Tracts in Theoretical Computer Science. Cambridge University Press, 2001. doi:10.1017/CBO9781107050884.
- [16] Francis Borceux. Handbook of Categorical Algebra, volume 3 of Encyclopedia of Mathematics and its Applications. Cambridge University Press, 1994.
- [17] Francis Borceux. Handbook of Categorical Algebra, volume 1 of Encyclopedia of Mathematics and its Applications. Cambridge University Press, 1994.
- [18] Francis Borceux and Dominique Dejean. Cauchy completion in category theory. Cahiers de Topologie et Géométrie Différentielle Catégoriques, 27(2):133–146, 1986. URL: http://www.numdam.org/item/?id=CTGDC_1986__27_2_133_0.
- [19] Milan Božić and Kosta Došen. Models for normal intuitionistic modal logics. Studia Logica, 43(3):217–245, 1984. doi:10.1007/BF02429840.
- [20] Marta Bunge. Categories of Set-Valued Functors. PhD thesis, University of Pennsylvania, 1966.
- [21] Gian Luca Cattani and Glynn Winskel. Presheaf models for concurrency. In Dirk Dalen and Marc Bezem, editors, Computer Science Logic, volume 1258 of Lecture Notes in Computer Science, pages 58–75, Berlin, Heidelberg, 1997. Springer Berlin Heidelberg. doi:10.1007/3-540-63172-0_32.
- [22] Gian Luca Cattani and Glynn Winskel. Profunctors, open maps and bisimulation. Mathematical Structures in Computer Science, 15(3):553–614, 2005. doi:10.1017/S0960129505004718.
- [23] Alexander Chagrov and Michael Zakharyaschev. Modal Logic. Number 35 in Oxford Logic Guides. Oxford University Press, 1996. doi:10.1093/oso/9780198537793.001.0001.
- [24] Denis-Charles Cisinski. Higher Categories and Homotopical Algebra. Cambridge University Press, 2019. doi:10.1017/9781108588737.
- [25] Ranald Clouston. Fitch-Style Modal Lambda Calculi. In Christel Baier and Ugo Dal Lago, editors, Foundations of Software Science and Computation Structures, volume 10803 of Lecture Notes in Computer Science, pages 258–275, Cham, 2018. Springer International Publishing. doi:10.1007/978-3-319-89366-2_14.
- [26] Roy L. Crole. Categories for Types. Cambridge University Press, 1993. doi:10.1017/CBO9781139172707.
- [27] Anupam Das and Sonia Marin. On intuitionistic diamonds (and lack thereof). In Revantha Ramanayake and Josef Urban, editors, Automated Reasoning with Analytic Tableaux and Related Methods, volume 14278 of Lecture Notes in Computer Science, pages 283–301. Springer Nature Switzerland, 2023. doi:10.1007/978-3-031-43513-3_16.
- [28] B. A. Davey and H. A. Priestley. Introduction to Lattices and Order. Cambridge University Press, 2nd edition, 2002. doi:10.1017/CBO9780511809088.
- [29] D.H.J. De Jongh and A.S. Troelstra. On the connection of partially ordered sets with some pseudo-boolean algebras. Indagationes Mathematicae (Proceedings), 69:317–329, 1966. doi:10.1016/S1385-7258(66)50036-1.
- [30] Wojciech Dzik, Jouni Järvinen, and Michiro Kondo. Intuitionistic propositional logic with Galois connections. Logic Journal of IGPL, 18(6):837–858, 2010. doi:10.1093/jigpal/jzp057.
- [31] M. Erné. The ABC of order and topology. In H. Herrlich and H. E. Porst, editors, Category Theory at Work, Research and Exposition in Mathematics, pages 57–83. Heldermann Verlag, 1991. URL: https://www.heldermann.de/R&E/RAE18/ctw05.pdf.
- [32] Leo Esakia. Heyting Algebras: Duality Theory, volume 50 of Trends in Logic. Springer International Publishing, 2019. doi:10.1007/978-3-030-12096-2.
- [33] W. B. Ewald. Intuitionistic Tense and Modal Logic. The Journal of Symbolic Logic, 51(1):166–179, 1986. doi:10.2307/2273953.
- [34] M. Fiore, G. Plotkin, and D. Turi. Abstract syntax and variable binding. In Proceedings. 14th Symposium on Logic in Computer Science (Cat. No. PR00158), pages 193–202, Trento, Italy, 1999. IEEE Computer Society Press. doi:10.1109/LICS.1999.782615.
- [35] Marcelo Fiore. Second-Order and Dependently-Sorted Abstract Syntax. In 2008 23rd Annual IEEE Symposium on Logic in Computer Science, pages 57–68. IEEE, 2008. doi:10.1109/LICS.2008.38.
- [36] Marcelo Fiore and Chung-Kil Hur. Second-Order Equational Logic (Extended Abstract). In Anuj Dawar and Helmut Veith, editors, Computer Science Logic, volume 6247 of Lecture Notes in Computer Science, pages 320–335. Springer Berlin Heidelberg, 2010. doi:10.1007/978-3-642-15205-4_26.
- [37] Marcelo Fiore and Ola Mahmoud. Second-Order Algebraic Theories: (Extended Abstract). In Petr Hliněnỳ and Antonín Kučera, editors, Mathematical Foundations of Computer Science 2010, volume 6281 of Lecture Notes in Computer Science, pages 368–380. Springer Berlin Heidelberg, 2010. doi:10.1007/978-3-642-15155-2_33.
- [38] Gisèle Fischer Servi. Semantics for a class of intuitionistic modal calculi. In Italian Studies in the Philosophy of Science, volume 47 of Boston Studies in the Philosophy of Science, pages 59–72. Springer Netherlands, 1980. doi:10.1007/978-94-009-8937-5_5.
- [39] Melvin Fitting. Intuitionistic Logic, Model Theory and Forcing. Studies in Logic and the Foundation of Mathematics. North-Holland, 1969.
- [40] Mai Gehrke. Stone duality, topological algebra, and recognition. Journal of Pure and Applied Algebra, 220(7):2711–2747, 2016. doi:10.1016/j.jpaa.2015.12.007.
- [41] Mai Gehrke and Sam van Gool. Topological Duality for Distributive Lattices: Theory and Applications. Number 61 in Cambridge Tracts in Theoretical Computer Science. Cambridge University Press, 2024. URL: http://arxiv.org/abs/2203.03286.
- [42] S. Ghilardi and G. C. Meloni. Modal and tense predicate logic: Models in presheaves and categorical conceptualization. In Francis Borceux, editor, Categorical Algebra and its Applications, volume 1348 of Lecture Notes in Mathematics, pages 130–142. Springer Berlin Heidelberg, 1988. doi:10.1007/BFb0081355.
- [43] Silvio Ghilardi. Presheaf semantics and independence results for some non-classical first-order logics. Archive for Mathematical Logic, 29(2):125–136, 1989. doi:10.1007/BF01620621.
- [44] Silvio Ghilardi. Incompleteness Results in Kripke Semantics. The Journal of Symbolic Logic, 56(2):517–538, 1991. doi:10.2307/2274697.
- [45] Robert Goldblatt. Metamathematics of Modal Logic. PhD Thesis, Victoria University, Wellington, 1974.
- [46] Robert Goldblatt. Mathematical modal logic: A view of its evolution. In Dov M. Gabbay, editor, Handbook of the History of Logic, volume 7. Elsevier, 2006. doi:10.1016/S1570-8683(03)00008-9.
- [47] Valentin Goranko and Antje Rumberg. Temporal Logic. In Edward N. Zalta and Uri Nodelman, editors, The Stanford Encyclopedia of Philosophy. Metaphysics Research Lab, Stanford University, Fall 2023 edition, 2023.
- [48] Daniel Gratzer. Normalization for Multimodal Type Theory. In Proceedings of the 37th Annual ACM/IEEE Symposium on Logic in Computer Science, pages 1–13. Association for Computing Machinery, 2022. doi:10.1145/3531130.3532398.
- [49] Daniel Gratzer. Syntax and semantics of modal type theory. PhD thesis, Aarhus University, 2023.
- [50] Daniel Gratzer, Evan Cavallo, G. A. Kavvos, Adrien Guatto, and Lars Birkedal. Modalities and Parametric Adjoints. ACM Transactions on Computational Logic, 23(3):1–29, 2022. doi:10.1145/3514241.
- [51] Daniel Gratzer, G. A. Kavvos, Andreas Nuyts, and Lars Birkedal. Multimodal Dependent Type Theory. In Proceedings of the 35th Annual ACM/IEEE Symposium on Logic in Computer Science, pages 492–506. Association for Computing Machinery, 2020. doi:10.1145/3373718.3394736.
- [52] Daniel Gratzer, G. A. Kavvos, Andreas Nuyts, and Lars Birkedal. Multimodal Dependent Type Theory. Logical Methods in Computer Science, 17(3), 2021. doi:10.46298/lmcs-17(3:11)2021.
- [53] Makoto Hamana. Free -Monoids: A Higher-Order Syntax with Metavariables. In Wei-Ngan Chin, editor, Programming Languages and Systems, volume 3302 of Lecture Notes in Computer Science, pages 348–363. Springer Berlin Heidelberg, 2004. doi:10.1007/978-3-540-30477-7_23.
- [54] Yasusi Hasimoto. Heyting Algebras with Operators. Mathematical Logic Quarterly, 47(2):187–196, 2001. doi:10.1002/1521-3870(200105)47:2<187::AID-MALQ187>3.0.CO;2-J.
- [55] Jens Hemelaer. Toposes over which essential implies locally connected. Cahiers de Topologie et Géométrie Différentielle Catégoriques, LXIII(4):425–451, 2022. arXiv:2204.02749.
- [56] Claudio Hermida. A categorical outlook on relational modalities and simulations. Information and Computation, 209(12):1505–1517, 2011. doi:10.1016/j.ic.2010.09.009.
- [57] Martin Hofmann. Syntax and Semantics of Dependent Types. In Andrew M. Pitts and P. Dybjer, editors, Semantics and Logics of Computation, pages 79–130. Cambridge University Press, 1997. URL: https://www.tcs.ifi.lmu.de/mitarbeiter/martin-hofmann/pdfs/syntaxandsemanticsof-dependenttypes.pdf, doi:10.1017/CBO9780511526619.004.
- [58] Martin Hofmann. Semantical analysis of higher-order abstract syntax. In Proceedings. 14th Symposium on Logic in Computer Science (Cat. No. PR00158), pages 204–213. IEEE Computer Society Press, 1999. doi:10.1109/LICS.1999.782616.
- [59] Michael Huth and Mark Ryan. Logic in Computer Science: Modelling and Reasoning about Systems. Cambridge University Press, 2nd edition, 2004.
- [60] Peter T. Johnstone. Open maps of toposes. Manuscripta Mathematica, 31(1):217–247, 1980. doi:10.1007/BF01303275.
- [61] Peter T. Johnstone. Stone Spaces. Number 3 in Cambridge Studies in Advanced Mathematics. Cambridge University Press, 1982.
- [62] Peter T. Johnstone. Sketches of an Elephant: A Topos Theory Compendium, volume 1. Oxford University Press, 2002.
- [63] Peter T. Johnstone. Sketches of an Elephant: A Topos Theory Compendium, volume 2. Oxford University Press, 2002.
- [64] Bjarni Jonnson and Alfred Tarski. Boolean algebras with operators. part II. American Journal of Mathematics, 74(1):127, 1952. doi:10.2307/2372074.
- [65] André Joyal, Mogens Nielsen, and Glynn Winskel. Bisimulation from Open Maps. Information and Computation, 127(2):164–185, 1996. doi:10.1006/inco.1996.0057.
- [66] G. Max Kelly. Basic concepts of enriched category theory. Reprints in Theory and Applications of Categories, (10):1–136, 2005. Originally published as: Cambridge University Press, Lecture Notes in Mathematics 64, 1982. URL: http://www.tac.mta.ca/tac/reprints/articles/10/tr10abs.html.
- [67] Kohei Kishida. Categories and modalities. In Elaine Landry, editor, Categories for the Working Philosopher. Oxford University Press, 2018. doi:10.1093/oso/9780198748991.003.0009.
- [68] Saul Kripke. Semantical Considerations on Modal Logic. Acta Philosophica Fennica, 16:83–94, 1963. doi:10.1002/malq.19630090502.
- [69] Saul A. Kripke. Semantical Analysis of Modal Logic I. Normal Modal Propositional Calculi. Zeitschrift für Mathematische Logik und Grundlagen der Mathematik, 9(5-6):67–96, 1963. doi:10.1002/malq.19630090502.
- [70] Saul A. Kripke. Semantical Analysis of Modal Logic II. Non-Normal Modal Propositional Calculi. In Alfred Tarski, Leon Henkin, and J. W. Addison, editors, The theory of models, Proceedings of the 1963 International Symposium at Berkeley, pages 206–220. North-Holland, 1965.
- [71] Joachim Lambek. From lambda calculus to cartesian closed categories. In Jonathan P. Seldin and J. Roger Hindley, editors, To H. B. Curry: Essays on Combinatory Logic, Lambda Calculus, and Formalism, pages 376–402. Academic Press, 1980.
- [72] Joachim Lambek and Philip J. Scott. Introduction to Higher-Order Categorical Logic. Number 7 in Cambridge Studies in Advanced Mathematics. Cambridge University Press, 1988.
- [73] F. William Lawvere. Metric spaces, generalized logic, and closed categories. Rendiconti del Seminario Matematico e Fisico di Milano, 43(1):135–166, 1973. doi:10.1007/BF02924844.
- [74] Fosco Loregian. (Co)end Calculus. Cambridge University Press, 2021. doi:10.1017/9781108778657.
- [75] Saunders Mac Lane. Categories for the Working Mathematician, volume 5 of Graduate Texts in Mathematics. Springer New York, 2 edition, 1978. doi:10.1007/978-1-4757-4721-8.
- [76] Saunders Mac Lane and Ieke Moerdijk. Sheaves in Geometry and Logic: A First Introduction to Topos Theory. Universitext. Springer New York, 1994. doi:10.1007/978-1-4612-0927-0.
- [77] Colin McLarty. Every Grothendieck topos has a one-way site. Theory and Applications of Categories, 16(5):123–126, 2006. URL: http://www.tac.mta.ca/tac/volumes/16/5/16-05abs.html.
- [78] M. Menni and C. Smith. Modes of adjointness. Journal of Philosophical Logic, 43(2):365–391, 2014. doi:10.1007/s10992-012-9266-y.
- [79] John C. Mitchell and Eugenio Moggi. Kripke-style models for typed lambda calculus. Annals of Pure and Applied Logic, 51(1-2):99–124, 1991. doi:10.1016/0168-0072(91)90067-V.
- [80] Mogens Nielsen, Gordon Plotkin, and Glynn Winskel. Petri nets, event structures and domains, Part I. Theoretical Computer Science, 13(1):85–108, 1981. doi:10.1016/0304-3975(81)90112-2.
- [81] F. J. Oles. Type Algebras, Functor Categories, and Block Structure. In Maurice Nivat and John C. Reynolds, editors, Algebraic Methods in Semantics, pages 543–573. Cambridge University Press, 1985.
- [82] Frank J. Oles. Functor Categories and Store Shapes. In Peter W. O’Hearn and Robert D. Tennent, editors, Algol-like Languages, volume 2, pages 3–12. Birkhäuser Boston, Boston, MA, 1997. doi:10.1007/978-1-4757-3851-3_1.
- [83] Ewa Orłowska and Ingrid Rewitzky. Discrete Dualities for Heyting algebras with Operators. Fundamenta Informaticae, 81(1):275–295, 2007.
- [84] Jorge Picado and Aleš Pultr. Frames and Locales: Topology without points. Frontiers in Mathematics. Springer Basel, 2012. doi:10.1007/978-3-0348-0154-6.
- [85] Gordon D. Plotkin and Colin Stirling. A framework for intuitionistic modal logics. In Joseph Y. Halpern, editor, Proceedings of the 1st Conference on Theoretical Aspects of Reasoning about Knowledge, Monterey, CA, USA, March 1986. Morgan Kaufmann, 1986.
- [86] George N. Raney. Completely distributive complete lattices. Proceedings of the American Mathematical Society, 3(5):677–680, 1952. doi:10.1090/S0002-9939-1952-0052392-3.
- [87] Emily Riehl. Categorical homotopy theory, volume 24 of New Mathematical Monographs. Cambridge University Press, 2014. doi:10.1017/CBO9781107261457.
- [88] Emily Riehl. Category Theory in Context. Dover Publications, 2016. URL: http://www.math.jhu.edu/~eriehl/context.pdf.
- [89] Mehrnoosh Sadrzadeh and Roy Dyckhoff. Positive logic with adjoint modalities: proof theory, semantics, and reasoning about information. The Review of Symbolic Logic, 3(3):351–373, 2010. doi:10.1017/S1755020310000134.
- [90] Davide Sangiorgi. On the origins of bisimulation and coinduction. ACM Transactions on Programming Languages and Systems, 31(4):1–41, 2009. doi:10.1145/1516507.1516510.
- [91] Krister Segerberg. Decidability of S4.1. Theoria, 34(1):7–20, 1968. doi:10.1111/j.1755-2567.1968.tb00335.x.
- [92] Alex K. Simpson. The Proof Theory and Semantics of Intuitionistic Modal Logic. PhD thesis, The University of Edinburgh, 1994. URL: http://hdl.handle.net/1842/407.
- [93] Morten Heine Sørensen and Pawel Urzyczyn. Lectures on the Curry-Howard Isomorphism. Elsevier, 2006.
- [94] V. H. Sotirov. Modal theories with intuitionistic logic. In Mathematical Logic: Proceedings of the Conference on Mathematical Logic, Dedicated to the Memory of A.A. Markov (1903-1979): Sofia, September 22-23, 1980. Bulgarian Academy of Sciences, 1984.
- [95] David Spivak. Conditions for a functor to induce a logical functor between presheaf toposes? MathOverflow, 2016. URL: https://mathoverflow.net/q/253878, arXiv:https://mathoverflow.net/q/253878.
- [96] S. K. Thomason. Categories of frames for modal logic. The Journal of Symbolic Logic, 40(3):439–442, 1975. doi:10.2307/2272167.
- [97] Nachiappan Valliappan, Fabian Ruch, and Carlos Tomé Cortiñas. Normalization for Fitch-style Modal Calculi. Proceedings of the ACM on Programming Languages, 6(ICFP):772–798, 2022. doi:10.1145/3547649.
- [98] Johan van Benthem. Modal logic for open minds. Number 199 in CSLI lecture notes. Center for the Study of Language and Information, Stanford, California, 2010.
- [99] Philip Wadler. Propositions as types. Communications of the ACM, 58(12):75–84, 2015. URL: http://homepages.inf.ed.ac.uk/wadler/papers/propositions-as-types/propositions-as-types.pdf, doi:10.1145/2699407.
- [100] Glynn Winskel. Prime algebraicity. Theoretical Computer Science, 410(41):4160–4168, 2009. doi:10.1016/j.tcs.2009.06.015.
- [101] F. Wolter and M. Zakharyaschev. The relation between intuitionistic and classical modal logics. Algebra and Logic, 36(2):73–92, 1997. doi:10.1007/BF02672476.
- [102] Frank Wolter and Michael Zakharyaschev. Intuitionistic Modal Logic. In Andrea Cantini, Ettore Casari, and Pierluigi Minari, editors, Logic and Foundations of Mathematics: Selected Contributed Papers of the Tenth International Congress of Logic, Methodology and Philosophy of Science, Florence, August 1995, pages 227–238. Springer Netherlands, Dordrecht, 1999. doi:10.1007/978-94-017-2109-7_17.
- [103] Frank Wolter and Michael Zakharyaschev. Intuitionistic Modal Logics as Fragments of Classical Bimodal Logics. In Ewa Orłowska, editor, Logic at Work: Essays Dedicated to the Memory of Helena Rasiowa, Studies in Fuzziness and Soft Computing, pages 168–186. Physica Heidelberg, 1999.
- [104] David Yetter. On right adjoints to exponential functors. Journal of Pure and Applied Algebra, 45(3):287–304, 1987. doi:10.1016/0022-4049(87)90077-6.