AI Glossary / Graphs and symbolic AI / Intermediate

Domain elements

Definition

Domain elements are the objects that belong to the domain of a logical model. They are what variables range over and what names or other terms can refer to. Depending on the model, elements might represent people, machines, numbers, events, or abstract entities; they are not necessarily the written symbols used to describe those things.

What it means

A logical statement needs a universe of things it is about. Saying every item is checked means little until the model determines which objects count as items under consideration. Domain elements provide that underlying population.
In first-order logic, the model also interprets predicate and function symbols. A predicate identifies which elements, or ordered combinations of elements, satisfy a relationship. A constant symbol names an element. Keeping these roles separate prevents confusion between a real modeled object, the label used for it, and the properties asserted about it.

How it works

Begin with a nonempty set of elements under ordinary classical first-order semantics. Assign each constant symbol an element of that set. Interpret a unary predicate as a set of elements, and a binary predicate as a set of ordered pairs. A variable assignment selects a domain element for each relevant variable.
A universal statement must hold for every possible element assigned to its quantified variable. An existential statement needs at least one suitable element. The domain may be finite or infinite. Its size does not have to equal the number of names in the language: some elements may be unnamed, and several names may designate one element unless additional axioms rule that out.

Three practical examples

• A machine-inspection model uses three physical machine objects as its domain and a predicate to identify those already checked.
• A mathematical model uses integers as domain elements, allowing a quantified formula to discuss numbers without listing them individually.
• A scheduling model includes people and rooms, with predicates distinguishing their roles and relationships connecting bookings to the relevant objects.

Think of it this way

Think of the objects in a labeled display case. The objects correspond to domain elements, labels to names, and a written description such as blue applies to selected objects. The analogy is limited because logical domains can be infinite, objects need not be physical, and an object can have multiple names or no name in the language.

Worked example

Let a model have domain D = {u,v,w}, where these letters are our descriptive labels for three elements. Let the logical constant MachineA refer to u and MachineB refer to v. Interpret Checked as the set {u,v}.
Then Checked(MachineA) is true, because MachineA denotes u and u belongs to the Checked set. The statement that some object is checked is also true. However, the statement that every object is checked is false, because w is a domain element outside that set. This remains so even if the logical language has no constant naming w.
Now add a constant AliasA that also denotes u. The domain still contains three elements; adding a name did not add an object. If we instead change the interpretation of Checked to {u,v,w}, the universal statement becomes true without changing the domain at all. The example separates three operations that software interfaces can blur: adding a name, adding an element, and changing a predicate's interpretation.

Illustrative case study

At fictional Elm Asset Models, two imported labels refer to the same machine. A prototype treats each label as a separate object and overcounts the modeled inventory. Reviewers establish an identity mapping before evaluating statements about all machines.
They also find an unnamed test fixture in the intended scope and decide explicitly whether it belongs in the model. This fictional case illustrates why identity and domain membership require deliberate decisions. It does not claim that formal semantics automatically resolves ambiguous records in a real asset database.

Common misconceptions and mistakes

A domain element is not the same as a variable, constant symbol, or data field. Those are parts of the representation or language. Their interpretations connect them to the objects in the model.
Do not assume that only named objects exist. Ordinary first-order models permit unnamed elements. Do not assume different names identify different objects either; unique-name assumptions must be justified or explicitly supplied. Finally, a domain in this sense is different from the set of allowable values for a particular constraint-satisfaction variable, although both uses involve sets of possible objects or values. Explain which meaning applies when moving between logic and optimization.

Business use

Businesses using logical rules should specify the modeled objects before writing broad statements about all of them. Decide how identity is established across systems, what belongs in scope, and whether different types of objects share one domain. Review how missing records affect conclusions. A rule proven true for one model depends on that model's elements and interpretations; it is not automatically a statement about every relevant object in the outside world.

People also ask

Must every element have a name?

No. A model may contain elements that no constant in the language names.

Can two constants refer to one element?

Yes, unless the theory or chosen semantics requires them to be distinct.

Are predicates domain elements?

In ordinary first-order semantics, predicate symbols are interpreted as relations over elements. They are not automatically objects that the language's variables range over.

Check your understanding

A model contains four elements but only three constant symbols. Does that make the model invalid?

Answer: No. Constants need not name every element, and the number of constants does not determine the domain size. Validity of particular statements depends on their interpretation and the theory's additional requirements.

Related terms

• Domain
• Interpretation
• Predicate (logic)
• Domain closure
• Unique name assumption
• constant symbol
• variable assignment
• predicate
• quantifier

Learn next

Extend the worked model with a binary relation Inspects containing the ordered pair (u,v). Evaluate Inspects(MachineA,MachineB) and its reversed form. Explain why knowing one ordered pair belongs to the relation does not establish that the reversed pair does too.

More in Graphs and symbolic AI

• Distributed representation - A distributed representation encodes an item through a pattern across several representational units, with individual units participating in the...
• Distributed tree search - Distributed tree search divides the exploration of a search tree among cooperating processors or workers.
• Divisive Communities - Divisive communities are groups identified by progressively splitting a graph into smaller parts.
• Dominant strategy - A dominant strategy is a strategy that gives a player an outcome at least as good as every alternative, whatever strategies the other players choose, under...
• Dominant strategy equilibrium - A dominant strategy equilibrium is a joint choice of strategies in which every player uses a dominant strategy.
• Double hashing - Double hashing is an open-addressing method for resolving collisions in a hash table.

Sources

Russell and Norvig, First-Order Logic, original author-hosted chapter, sections on models, domain elements, and interpretations: https://aima.cs.berkeley.edu/4th-ed/pdfs/newchap08.pdf Stanford knowledge-representation course, original examples distinguishing unique names and domain closure: https://www-formal.stanford.edu/leora/krcourse/homeworks/hw2sampleaxioms.html

Previous: Domain driven data mining Next: Domain generalization