🔢 mathematics

Free constructions for comprehension categories

This paper investigates the relationship between Jacobs comprehension categories and the subclass of Lawvere-Ehrhard comprehension categories by characterizing the latter through term and type morphism fibrations, and subsequently providing constructions for free comprehension categories over fibrations and free Lawvere-Ehrhard comprehension categories over Jacobs comprehension categories.

Original authors: Francesco Dagnino, Jacopo Emmenegger, Andrea Giusto

Published 2026-07-30

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Francesco Dagnino, Jacopo Emmenegger, Andrea Giusto

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are building a massive, interlocking Lego castle. In the world of computer science, specifically in a field called "type theory," these bricks are called "types," and the instructions for how they fit together are the rules of a programming language. Just like in real life, if you try to stack a heavy stone on a flimsy plastic piece, the whole thing collapses. To prevent this, computer scientists use "types" to make sure code is safe and logical. But sometimes, the rules get complicated. What if you want to say that a "dog" is also a "mammal"? Or that a "red ball" is a specific kind of "ball"? This is where things get tricky.

To handle these complex relationships, mathematicians and computer scientists use a tool called "category theory." Think of this as a super-powerful map that doesn't just show where the Lego bricks are, but how they can be transformed into one another. One popular way to draw this map is using something called a "fibration." If you imagine a stack of transparent sheets, a fibration is like a way of organizing those sheets so that if you slide one sheet (a "context" or a set of rules), the shapes drawn on it (the "types") move along with it perfectly. This paper dives deep into two different ways of drawing these maps, trying to figure out which one is better and how to turn one into the other.

The paper, titled "Free Constructions for Comprehension Categories," is written by Francesco Dagnino, Jacopo Emmenegger, and Andrea Giusto. It tackles a specific puzzle in the world of type theory: the relationship between two different models called "Jacobs comprehension categories" and "Lawvere-Ehrhard comprehension categories."

Think of a Jacobs comprehension category as a very flexible, open-ended workshop. In this workshop, you have your Lego bricks (types) and your instructions (contexts). You also have a special rulebook that tells you how to extend your instructions by adding a new variable, like saying "let's add a variable x of type A." In this model, the "morphisms" (which are like the rules for turning one type into another, or "subtyping") are treated as separate, independent pieces of data. It's like having a box of extra connectors that you can use to link bricks, but they aren't strictly tied to the bricks themselves. This makes the model very general, but sometimes a bit wild and hard to control because there are so many ways to connect things.

On the other hand, the paper introduces Lawvere-Ehrhard comprehension categories as a more disciplined, "tamed" version of the workshop. In this stricter model, the connection between types isn't just a loose connector; it's built into the very fabric of the system. The authors show that in a Lawvere-Ehrhard world, every "term" (a specific instance of a type, like a specific dog) is completely determined by a special kind of "type morphism" coming from a "unit type" (think of this as a generic "thing" or a universal placeholder). It's as if every specific Lego figure you build is automatically defined by how it relates to a single, master "generic" figure. This creates a tighter, more predictable relationship between the rules and the objects.

The main discovery of the paper is that these two models aren't enemies; they are related in a very specific, mathematical way. The authors prove that Lawvere-Ehrhard categories are essentially Jacobs categories where the "morphisms" (the connectors) and the "terms" (the specific figures) are perfectly matched up, like two sides of the same coin. They show that if you have a Jacobs category where every type has a unique "unit" connection, it automatically becomes a Lawvere-Ehrhard category.

But the real magic of the paper lies in the "free constructions." The authors don't just compare the two; they build a machine that can turn one into the other. They describe three step-by-step processes:

From Fibration to Jacobs: They show how to take a basic fibration (just a stack of sheets) and automatically build a full Jacobs comprehension category on top of it. This is like taking a pile of raw Lego bricks and automatically generating a full instruction manual for how to extend them.
From Jacobs to "Terminals": They show how to take a Jacobs category and add "fibred terminal objects." In our Lego analogy, this is like adding a special "universal baseplate" to every single instruction set, ensuring that every context has a unique, standard starting point.
From "Terminals" to Lawvere-Ehrhard: Finally, they show how to take that enhanced Jacobs category and force it to become a Lawvere-Ehrhard category. This step is the most complex; it involves identifying and merging different "connectors" that were doing the same job, effectively cleaning up the workshop so that every connection is unique and necessary.

The authors are very sure of their results. They don't just suggest these connections; they provide rigorous mathematical proofs (using things called "2-adjunctions" and "coequalizers") that these constructions work perfectly. They demonstrate that you can start with a simple fibration and, by applying these three steps in order, you will always end up with a Lawvere-Ehrhard comprehension category.

Why does this matter? Because in the world of programming languages, having a "proof-relevant" subtyping system (where different ways of converting types matter) is becoming increasingly important. This paper gives computer scientists the tools to build these complex systems from scratch, ensuring that the rules they create are consistent and mathematically sound. It's like giving architects a set of blueprints that guarantees their skyscrapers won't collapse, no matter how many new floors they add. The paper concludes by suggesting that these "free constructions" could be the key to building new, more powerful programming languages that handle complex type relationships with ease.

Technical Summary: Free Constructions for Comprehension Categories

Problem Statement
The paper addresses the structural relationship between two categorical frameworks for modeling dependent type theories: Jacobs Comprehension Categories (JCCs) and Lawvere-Ehrhard Comprehension Categories (LECCs). While JCCs provide a general unifying framework for type dependency (subsuming models like categories with families and contextual categories), they treat type morphisms as independent data. In contrast, LECCs (originally D-categories) impose a stricter universal property where extended contexts are characterized by type morphisms from a unit type.

Recent developments in dependent type theory have highlighted the importance of non-trivial type morphisms for modeling subtyping and proof-relevant coercions. However, the precise structural distinction between JCCs and LECCs, particularly regarding how they handle type morphisms and terms, requires clarification. Furthermore, there is a lack of systematic "free constructions" to generate LECCs from arbitrary fibrations or to transform JCCs into LECCs, which would allow for the modular extension of models with specific features like unit types and proof-irrelevance.

Methodology
The authors employ a categorical approach based on Grothendieck fibrations and 2-category theory. The methodology proceeds in three main stages:

Characterization: The paper first establishes a precise characterization of LECCs within the broader class of JCCs. It compares two associated fibrations: the fibration of terms ( $\dot{p}$ ) and the fibration of global type morphisms ( $\tilde{p}$ ). The authors prove that a JCC with fibred terminal objects is an LECC if and only if these two fibrations are isomorphic. This isomorphism implies that terms are completely determined by type morphisms from the unit type.
Free Constructions: The paper constructs three universal (free) 2-functors to relate these structures:
- $F_J$ : Constructs the free JCC over an arbitrary fibration. This corresponds to adding context extension rules (weakening and variables) to a theory with only substitution.
- $F_T$ : Constructs the free JCC with fibred terminal objects (preserved by the comprehension functor) over a given JCC. This corresponds to adding a unit type $\top$ inhabited by a unique term in every context.
- $F_{LE}$ : Constructs the free LECC over a JCC with fibred terminal objects. This is achieved via a 2-coequalizer construction that forces an isomorphism between terms and global type morphisms, effectively identifying type morphisms that correspond to the same term.
Adjunctions: The authors prove that these constructions form left bi-adjoints (or left 2-adjoints in the case of $F_{LE}$ ) to the respective forgetful functors. This establishes the universal property of the constructions, ensuring they are the most general way to add the required structure without imposing unnecessary constraints.

Key Contributions and Results

Structural Characterization: The paper identifies that the core difference between JCCs and LECCs lies in the treatment of type morphisms. In JCCs, they are independent; in LECCs, they are intrinsically linked to terms via the unit type. Theorem 3.16 provides the necessary and sufficient conditions for a JCC to be an LECC: the existence of an isomorphism between the fibration of terms and the fibration of global type morphisms.
Faithfulness and Proof-Irelevance: The authors analyze the case of faithful fibrations (where fibres are preorders). They prove that for LECCs, faithfulness is equivalent to proof-irrelevance (every type has at most one inhabitant). This equivalence does not hold for general JCCs, highlighting a specific rigidity in LECCs.
Universal Constructions:
- Theorem 4.9: Establishes that the 2-functor $F_J$ is left bi-adjoint to the forgetful functor $U_J: \mathbf{JCC} \to \mathbf{Fib}$ .
- Theorem 5.4: Establishes that $F_T$ is left bi-adjoint to $U_T: \mathbf{JCCT} \to \mathbf{JCC}$ , freely adding fibred terminal objects.
- Theorem 5.9: Establishes that $F_{LE}$ is left 2-adjoint to $U_{LE}: \mathbf{LECC} \to \mathbf{JCCT}$ . This construction uses a 2-coequalizer to enforce the bijection between terms and global type morphisms.
Composition: By composing these functors ( $F_{LE} \circ F_T \circ F_J$ ), the paper provides a direct construction of the free LECC over an arbitrary fibration.

Significance and Claims
The paper claims that these free constructions capture, in a principled categorical manner, the syntactic structures that differentiate these models. Specifically:

They provide a modular method for extending models with new features (e.g., adding a unit type or enforcing proof-irrelevance) without redefining the entire model from scratch.
They generate a wide variety of new, free examples of comprehension categories.
The analysis clarifies how the presence of non-trivial type morphisms yields a richer interaction among components of a dependent type theory, and how the LECC condition "tames" this interplay by making terms dependent on type morphisms.

The authors note that while the constructions are universal, they do not claim to solve the problem of how these structures interact with specific type constructors (like $\Pi$ -types or $\Sigma$ -types) in all cases, identifying this as a direction for future work. Similarly, the paper does not claim that the resulting bi-adjunctions are 2-monadic, though it suggests this as a potential avenue for future investigation to treat these structures as algebraic concepts.

More like this