In mathematics, logic, and computer science, a type theory is any of a class of formal systems, some of which can serve as alternatives to set theory as a foundation for all mathematics. In type theory, every "term" has a "type" and operations are restricted to terms of a certain type.
Type theory is closely related to (and in some cases overlaps with) type systems, which are a programming language feature used to reduce bugs. Type theory was created to avoid paradoxes in formal logics and rewrite systems.
Between 1902 and 1908 Bertrand Russell proposed various "theories of type" in response to his discovery that Gottlob Frege's version of naive set theory was afflicted with Russell's paradox. By 1908 Russell arrived at a "ramified" theory of types together with an "axiom of reducibility" both of which featured prominently in Whitehead and Russell's Principia Mathematica published between 1910 and 1913. They attempted to resolve Russell's paradox by first creating a hierarchy of types, then assigning each concrete mathematical (and possibly other) entity to a type. Entities of a given type are built exclusively from entities of those types that are lower in their hierarchy, thus preventing an entity from being assigned to itself. In the 1920s, Leon Chwistek and Frank P. Ramsey proposed an unramified type theory, now known as the "theory of simple types" or "simple type theory", which collapsed the hierarchy of the types in the earlier ramified theory and as such did not require the axiom of reducibility.
The common usage of "type theory" is when those types are used with a term rewrite system. The most famous early example is Alonzo Church's simply typed lambda calculus. Church's theory of types helped the formal system avoid the Kleene-Rosser paradox that afflicted the original untyped lambda calculus. Church demonstrated that it could serve as a foundation of mathematics and it was referred to as a higher-order logic.
Some other type theories include Per Martin-Löf's intuitionistic type theory, which has been the foundation used in some areas of constructive mathematics. Thierry Coquand's calculus of constructions and its derivatives are the foundation used by Coq, Lean, and others. The field is an area of active research, as demonstrated by homotopy type theory.
In a system of type theory, each term has a type. For example, , , and are all separate terms with the type for natural numbers. Traditionally, the term is followed by a colon and its type, such as - this means that the number is of type .
Type theories have explicit computation and it is encoded in rules for rewriting terms. These are called conversion rules or, if the rule only works in one direction, a reduction rule. For example, and are syntactically different terms, but the former reduces to the latter. This reduction is written .
Functions in type theory have a special reduction rule: the argument to the function is substituted for every instance of the parameter in the function definition. Let's say the function is defined as (using Church's lambda notation) or (using a more modern notation). Then, the function call would be reduced by substituting for every in the function definition. Thus, .
The type of a function is denoted with an arrow from the parameter type to the function's return type. Thus, . Calling or "applying" a function to an argument may be written with or without parentheses, so or . Omitting parentheses is more common, because multiple argument functions can be defined using currying.
There are many different set theories and many different systems of type theory, so what follows are generalizations.
The term reduces to . Since cannot be reduced further, it is called a normal form. A system of type theory is said to be strongly normalizing if all terms have a normal form and any order of reductions reaches it. Weakly normalizing systems have a normal form but some orders of reductions may loop forever and never reach it.
For a normalizing system, some borrow the word element from set theory and use it to refer to all closed terms that can reduce to the same normal form. A closed term has no parameters. (A term like with its parameter is called an open term.) Thus, and may be different terms but they are both from the element .
Convertibility is a similar idea which works for both open and closed terms. Two terms are convertible if they can be reduced to the same term. Confluent and weakly normalizing systems can test if two terms are convertible by checking if they both reduce to the same normal form.
A dependent type is a type that depends on a term or another type. Thus, the type returned by a function may depend on the argument to the function.
For example, a list of s of length 4 may be a different type than a list of s of length 5. In a type theory with dependent types, it is possible to define a function that takes a parameter "n" and returns a list containing "n" zeros. Calling the function with 4 would produce a term with a different type than if the function was called with 5.
Many systems of type theory have a type that represents equality of types and of terms. This type is different from convertibility, and is often denoted propositional equality.
In intuitionistic type theory, the equality type (also called the identity type) is known as for identity. There is a type when is a type and and are both terms of type . A term of type is interpreted as meaning that is equal to .
In practice, it is possible to build a type but there will not exist a term of that type. In intuitionistic type theory, new terms of equality start with reflexivity. If is a term of type , then there exists a term of type . More complicated equalities can be created by creating a reflexive term and then doing a reduction on one side. So if is a term of type , then there is a term of type and, by reduction, generate a term of type . Thus, in this system, the equality type denotes that two values of the same type are convertible by reductions.
Having a type for equality is important because it can be manipulated inside the system. There is usually no judgement to say two terms are not equal; instead, as in the Brouwer-Heyting-Kolmogorov interpretation, we map to , where is the bottom type having no values. There exists a term with type but not one of type .
A system of type theory requires some basic terms and types to operate on. Some systems build them out of functions using Church encoding. Other systems have inductive types: a set of base types and a set of type constructors that generate types with well-behaved properties. For example, certain recursive functions called on inductive types are guaranteed to terminate.
Induction-induction is a feature for declaring an inductive type and a family of types which depends on the inductive type.
Induction recursion allows a wider range of well-behaved types, allowing the type and recursive functions operating on it to be defined at the same time.
Types were created to prevent paradoxes, such as Russell's paradox. However, the motives that lead to those paradoxes--being able to say things about all types--still exist. So, many type theories have a "universe type", which contains all other types (and not itself).
In systems where you might want to say something about universe types, there is a hierarchy of universe types, each containing the one below it in the hierarchy. The hierarchy is defined as being infinite, but statements must only refer to a finite number of universe levels.
Type universes are particularly tricky in type theory. The initial proposal of intuitionistic type theory suffered from Girard's paradox.
Many systems of type theory, such as the simply-typed lambda calculus, intuitionistic type theory, and the calculus of constructions, are also programming languages. That is, they are said to have a "computational component". The computation is the reduction of terms of the language using rewriting rules.
A system of type theory that has a well-behaved computational component also has a simple connection to constructive mathematics through the BHK interpretation.
Non-constructive mathematics in these systems is possible by adding operators on continuations such as call with current continuation. However, these operators tend to break desirable properties such as canonicity and parametricity.
There is extensive overlap and interaction between the fields of type theory and type systems. Type systems are a programming language feature designed to identify bugs. Any static program analysis, such as the type checking algorithms in the semantic analysis phase of compiler, has a connection to type theory.
A prime example is Agda, a programming language which uses UTT (Lao's Unified Theory of dependent Types) for its type system. The programming language ML was developed for manipulating type theories (see LCF) and its own type system was heavily influenced by them.
This section needs expansion. You can help by adding to it. (May 2008)
The first computer proof assistant, called Automath, used type theory to encode mathematics on a computer. Martin-Löf specifically developed intuitionistic type theory to encode all mathematics to serve as a new foundation for mathematics. There is ongoing research into mathematical foundations using homotopy type theory.
Mathematicians working in category theory already had difficulty working with the widely accepted foundation of Zermelo-Fraenkel set theory. This led to proposals such as Lawvere's Elementary Theory of the Category of Sets (ETCS). Homotopy type theory continues in this line using type theory. Researchers are exploring connections between dependent types (especially the identity type) and algebraic topology (specifically homotopy).
Much of the current research into type theory is driven by proof checkers, interactive proof assistants, and automated theorem provers. Most of these systems use a type theory as the mathematical foundation for encoding proofs, which is not surprising, given the close connection between type theory and programming languages:
Type theory is also widely used in formal theories of semantics of natural languages, especially Montague grammar and its descendants. In particular, categorial grammars and pregroup grammars extensively use type constructors to define the types (noun, verb, etc.) of words.
The most common construction takes the basic types and for individuals and truth-values, respectively, and defines the set of types recursively as follows:
A complex type is the type of functions from entities of type to entities of type . Thus one has types like which are interpreted as elements of the set of functions from entities to truth-values, i.e. indicator functions of sets of entities. An expression of type is a function from sets of entities to truth-values, i.e. a (indicator function of a) set of sets. This latter type is standardly taken to be the type of natural language quantifiers, like everybody or nobody (Montague 1973, Barwise and Cooper 1981).
Although the initial motivation for category theory was far removed from foundationalism, the two fields turned out to have deep connections. As John Lane Bell writes: "In fact categories can themselves be viewed as type theories of a certain kind; this fact alone indicates that type theory is much more closely related to category theory than it is to set theory." In brief, a category can be viewed as a type theory by regarding its objects as types (or sorts), i.e. "Roughly speaking, a category may be thought of as a type theory shorn of its syntax." A number of significant results follow in this way:
The interplay, known as categorical logic, has been a subject of active research since then; see the monograph of Jacobs (1999) for instance.