Applied Category Theory

Applied category theory is the branch of mathematics that takes the abstract machinery of categories—objects, arrows, composition—and uses it to model, reason about, and compute with real systems: databases, quantum protocols, dynamical systems, networks, programming languages, and logical theories.

10 frameworks, from 1969 to today · Category Theory · Mathematics

still a working view ended · each leads to its entry below

Overview

Applied category theory is the branch of mathematics that takes the abstract machinery of categories—objects, arrows, composition—and uses it to model, reason about, and compute with real systems: databases, quantum protocols, dynamical systems, networks, programming languages, and logical theories1. Its practitioners believe that the same structural patterns recur across wildly different domains, and that a precise language for structure pays for itself when you need to translate between domains, compose systems, or prove that two constructions behave the same way. What is at stake is not merely elegance: category theory offers tools for compositionality—building complex things from simple parts while keeping track of how the parts fit together—and for correctness, since commutative diagrams and universal properties are checkable certificates that a construction does what it claims.

From pure foundations to applied ambition (1940s–1980s)

Category theory was born in 1945, when Samuel Eilenberg and Saunders Mac Lane introduced categories, functors, and natural transformations to make precise the sense in which different cohomology theories are "the same." For decades it lived inside algebraic topology, homological algebra, and algebraic geometry. Alexander Grothendieck's school in the late 1950s and 1960s used it to rebuild geometry on sheaves and schemes; William Lawvere's 1963 thesis and the topos-theoretic work that followed showed that category theory could serve as a foundation for mathematics itself. F. William Lawvere and Stephen Schanuel's work on categories of sets and on the "category of categories" as a foundation, and Lawvere's 1963 functorial semantics of algebraic theories, planted the idea that a theory is a category and a model is a functor into a category of sets.

The first sustained applications outside pure mathematics came from logic and computer science. In the late 1960s and 1970s, Lawvere's algebraic theories and the Curry–Howard correspondence—the observation that proofs in a formal system correspond to programs and propositions to types—gave categorical logic a computational reading. Joachim Lambek's work in the 1970s on cartesian closed categories and the lambda calculus made the correspondence precise: the simply typed lambda calculus is the internal language of a cartesian closed category. Dana Scott's 1969–1972 construction of models of the lambda calculus in continuous lattices, and his later domain theory, gave denotational semantics a categorical footing. By the 1980s, Eugenio Moggi's monadic semantics for computational effects (published 1991) and the rise of typed functional languages such as ML and Haskell made category theory a working tool for programming-language designers.

Categorical logic and type theory

The logic strand matured through the 1970s and 1980s. Lawvere's hyperdoctrines (1969–1970) gave a categorical account of first-order logic; Michael Makkai and Gonzalo Reyes's First Order Categorical Logic (1977) developed it systematically. Per Martin-Löf's dependent type theory (from 1972 onward) found categorical semantics in locally cartesian closed categories, and in the 1980s and 1990s the correspondence was extended to identity types and higher structure. The practical payoff is concrete: a type theory can be read as a category, so a proof assistant implementing that type theory is, in effect, a machine for constructing arrows in a category. This is the ancestry of modern proof assistants such as Coq and Agda, and of the later homotopy type theory program (Voevodsky and collaborators, around 2006–2013), in which types are interpreted as spaces and identity proofs as paths.

Monoidal categories, string diagrams, and quantum mechanics

A second strand concerns processes rather than theories. A monoidal category is a category equipped with a way to put objects side by side and arrows in parallel, with coherence conditions making the result associative and unital up to isomorphism2. In the 1980s and 1990s, André Joyal and Ross Street developed the theory of braided and tortile monoidal categories; the graphical calculus of string diagrams—pictures in which objects are wires and arrows are boxes—became a rigorous notation, with Joyal and Street's 1991 paper giving the formal correspondence between diagrams and morphisms.

This machinery turned out to fit quantum mechanics remarkably well. In the 2000s, Samson Abramsky and Bob Coecke developed categorical quantum mechanics, treating a physical process as a morphism in a dagger compact closed category and a composite system as a tensor product. The resulting graphical language—now often called the ZX-calculus, developed by Coecke and Ross Duncan around 2008—lets one reason about quantum protocols and circuits by rewriting diagrams. The same framework underlies the "quantum picturalism" program and has been used to design and verify quantum error-correcting codes and measurement-based quantum computation.

Databases, ontologies, and functorial data migration

A third strand applies category theory to information. In 1994, Peter Buneman, Leonid Libkin, and colleagues observed that database schemas and their instances could be modeled as categories and functors; a schema is a small category whose objects are tables and whose arrows are foreign-key references, and an instance is a functor from that schema to the category of sets. In the 2000s and 2010s, David Spivak developed this into functorial data migration: a schema mapping is a functor, and migrating data along it is given by left and right adjoints (the "sigma" and "pi" operations), with a third operation, "delta," given by precomposition. The framework, presented in Spivak's Category Theory for the Sciences (2014) and in work with Patrick Schultz and Ryan Wisnesky, gives a precise account of when one can translate data between schemas and what information is lost. It has been implemented in tools such as Wisnesky's Algebraic Query Language (AQL).

Operads, networks, and compositional systems

The idea that complex systems are built by plugging components into interfaces has an algebraic home in operads3. An operad is a structure whose operations take several inputs and produce one output, with a rule for substituting one operation into another's input—a formal grammar of composition3. Operads arose in algebraic topology in the 1960s and 1970s (J. Peter May's The Geometry of Iterated Loop Spaces, 1972, and the work of Boardman and Vogt), and were later applied to string theory and to the combinatorics of trees. In the 2010s, Spivak and others used operads and their colored variants to model networks—electrical circuits, chemical reaction networks, and dynamical systems—as algebras over an operad, so that composing networks is composing operations. This is the core of categorical systems theory: a dynamical system is a morphism in a category of systems, and wiring diagrams are the morphisms of an operad or a symmetric monoidal category.

Enriched and higher-categorical modeling

As applications grew, so did the need for categories whose hom-sets carry extra structure. Enriched category theory, developed by Lawvere (1973) and others, replaces the set of arrows between two objects with an object of a monoidal category—for instance, a metric space of distances, or a truth-value lattice of propositions. This lets one model graded or fuzzy relationships. Higher category theory, in which there are arrows between arrows and so on, entered applied work through the need to track coherence: when a construction is only associative up to a specified isomorphism, one needs the higher cells to record how those isomorphisms compose. The homotopy hypothesis—that higher groupoids model homotopy types—and the development of quasicategories by André Joyal (1984) and Jacob Lurie (from 2006) gave applied workers a toolkit for reasoning about systems up to homotopy, and for treating type theory, topology, and computation in one framework.

The present: a field with its own institutions

By the 2010s, applied category theory had become a recognizable field with its own venues. The Applied Category Theory conference series began in 2018 (the first at the Lorentz Center in Leiden), and the journal Compositionality was founded in 2019. Textbooks—Spivak's Category Theory for the Sciences (2014), Fong and Spivak's An Invitation to Applied Category Theory (2019), and Baez and Biamonte's Quantum Techniques in Stochastic Mechanics (2018)—made the material teachable. The central open questions are compositional: how to give a uniform semantics for systems that combine discrete and continuous behavior; how to make categorical constructions computationally effective at scale; and how far the higher-categorical and homotopical machinery can be pushed into engineering practice. The field's practitioners today work in mathematics, computer science, physics, linguistics, and biology, united by the conviction that the right diagram is often the shortest proof.

References

  1. Applied category theory, Wikipedia
  2. Monoidal category, Wikipedia
  3. Operad, Wikipedia

The frameworks of Applied Category Theory

Each framework is a way of working the field has used: when it held, and the question it answers.

  1. 1969-Present

    Categorical logic

    How does a logical theory become a category?

    Fly to its tower in Nousfarer

  2. 1969-Present

    Denotational semantics

    What mathematical object does a program denote?

    Fly to its tower in Nousfarer

  3. 1973-Present

    Enriched category

    What if arrows between objects carry extra structure?

    Fly to its tower in Nousfarer

  4. 1984-Present

    Higher category theory

    How do we track arrows between arrows?

    Fly to its tower in Nousfarer

  5. 1991-Present

    Monoidal and String-Diagram Calculus

    How do we draw and rewrite composed processes?

    Fly to its tower in Nousfarer

  6. 1994-Present

    Functorial Data Migration

    How do we translate data between schemas?

    Fly to its tower in Nousfarer

  7. 2004-Present

    Categorical Quantum Mechanics

    What is the diagrammatic language of quantum processes?

    Fly to its tower in Nousfarer

  8. 2010-Present

    Categorical Systems Theory

    How do we compose dynamical systems from parts?

    Fly to its tower in Nousfarer

  9. 2014-Present

    Enriched and higher-categorical modeling

    How do we model systems with graded or coherent structure?

    Fly to its tower in Nousfarer

  10. 2015-Present

    Operadic and network composition

    How do we plug components into interfaces formally?

    Fly to its tower in Nousfarer

How the ideas connect

Each arrow runs from an earlier idea to the later one it shaped. The list below says what passed between them.

Categorical logicDenotational semanticsEnriched categoryHigher category theoryMonoidal and String-Diagram CalculusFunctorial Data MigrationCategorical Quantum MechanicsCategorical Systems TheoryEnriched and higher-categorical model…Operadic and network composition

Beyond the field

Where Applied Category Theory's ideas came from in other fields, and what they led to there: the strongest 12 of 37 links.

A city in Nousfarer: a red-carpeted stair climbing terraces to a crowned tower on an acropolis, more towers round it, forest and snowy mountains beyond.
A city in Nousfarer, the game played in the browser.

Applied Category Theory is a planet in Nousfarer

Its city has a tower for each of these 10 frameworks. Land, walk in, and climb one concept at a time.

Land on Applied Category Theory

Related fields

Fields its ideas are linked to

More in Category Theory

See where Applied Category Theory sits on the Map of Human Knowledge