AI Glossary / Data engineering and architecture / Beginner

Identification (information)

Definition

Identification, in an information-management context, is the ability to distinguish and refer to a particular entity or information item without ambiguity within a defined scope. An identifier supplies the reference, such as a document ID or a database key. Reliable identification makes it possible to retrieve, connect, update or discuss the intended item rather than another item with similar descriptive details.

What it means

A person's name, a document title or a product description can help people recognize something, but these descriptions may repeat or change. A managed identifier provides a more controlled reference. Its scope matters: order 42 in one system need not be order 42 in another. Identification also differs from authentication, which checks an identity claim, and authorization, which determines permitted actions. Knowing the correct record identifier does not establish permission to access the record.

How it works

Designers decide what kind of thing receives an identifier and which authority issues it. They specify uniqueness, format, lifetime and rules for replacement or merging. Systems store the identifier and use it in references and relationships. When information moves between systems, a mapping may preserve the link between local identifiers. For published resources, persistent web identifiers can support continuing reference, but persistence requires operational maintenance. A string that looks unique is not a substitute for an agreed identification policy.

Three practical examples

• A glossary assigns each term a stable ID so editors can revise its wording without losing its completion history.
• A customer platform distinguishes two people who share the same name and avoids using a display label as a record key.
• An AI dataset records source-document IDs with extracted passages so reviewers can trace a model's evidence to the intended document version.

Think of it this way

Think of a coat-check ticket. The ticket number helps the attendant select the intended item even when many coats look alike. The number works because it belongs to a particular ticketing arrangement. The analogy has limits: information systems often keep references for years, combine collections and distinguish versions, while a temporary coat-check number may be safely reused after the event.

Worked example

Suppose System A contains invoice 104 for supplier North and System B contains invoice 104 for supplier South. Merging on the number 104 alone incorrectly suggests one invoice. Keeping the source system with the local number gives two references: A:104 and B:104. If later evidence shows that two differently numbered records represent the same invoice, a controlled mapping can link them to a shared entity identifier. That is a separate reconciliation decision; changing the syntax of the identifiers does not establish that two records describe the same real-world event.

Illustrative case study

In a fictional project, Larch Learning imports course materials from three departments. Several files are called Introduction, and filenames change during editing. The team assigns stable document IDs, keeps the original departmental references and records revisions separately. Reviewers can now discuss a specific item without relying on a filename. They still inspect apparent duplicates before merging them. The case illustrates a design approach, not a measured improvement, and it does not imply that introducing IDs automatically resolves conflicting content.

Common misconceptions and mistakes

Uniqueness inside one table is not global uniqueness. Identification is not the same as proving that the stored description is accurate. A single real-world entity may accidentally have several records, while an incorrect merge may place several entities under one identifier. A URL can change, so persistence needs planning. Sequential IDs are not secrets and should not be treated as access controls. Finally, assigning new identifiers during every import can destroy continuity unless the process preserves a reliable mapping to earlier references.

Business use

Organizations benefit when teams agree on the unit being identified: a person, account, subscription, document or document version. Those are not interchangeable. Document how IDs are issued and what happens after correction, retirement or merging. For AI-assisted workflows, retain identifiers through extraction, labeling and review so errors can be traced and corrected. Monitor unexpected collisions and broken references. Before linking datasets, assess whether the proposed linkage is appropriate and authorized; technical matching ability does not by itself justify joining personal information.

People also ask

Must an identifier be a number?

No. It may be a number, string, composite value or managed URI, depending on the system.

Does the same identifier always mean the same object?

Only within the relevant scheme and its rules. Context and issuing authority matter.

Can descriptive fields serve as identifiers?

Sometimes, if they satisfy the required uniqueness and stability rules, but repeated names and changing descriptions often make poor choices.

Check your understanding

Two databases each contain a customer with local ID 57. Is that enough evidence that they describe the same person?

Answer: No. The IDs may belong to independent namespaces. A documented cross-system mapping or an appropriate reconciliation process is needed before treating the records as one entity.

Related terms

• Primary key
• Referential integrity
• Identifier
• namespace
• entity resolution
• persistent identifier
• authentication

Learn next

Write an identification policy for a small collection: define what receives an ID, who issues it and whether it can ever be reused. Then practice combining two collections with overlapping local numbers. Compare identifying a document with identifying a particular revision of that document.

More in Data engineering and architecture

• Hybrid cloud storage - Hybrid cloud storage combines storage in an organization's local environment with cloud storage as part of a connected operating arrangement.
• Hypercube (communication pattern) - A hypercube communication pattern organizes exchanges between computing processes as if the processes occupied the corners of a multidimensional cube.
• Hyperjacking - Hyperjacking is malicious control or subversion of a hypervisor, the software layer that manages virtual machines.
• Identifier Domain - An identifier domain is the defined set and representation of values used to identify entities or records in a particular context.
• Identity column - An identity column is a database column whose values can be generated automatically when new rows are inserted.
• Imageability - Imageability is the capacity of something to evoke a mental image. The meaning depends on context.

Sources

Sources: W3C, “Data on the Web Best Practices,” https://www.w3.org/TR/dwbp/; W3C, “Best Practices for Publishing Linked Data,” https://www.w3.org/TR/ld-bp/; PostgreSQL, “Constraints,” https://www.postgresql.org/docs/18/ddl-constraints.html. W3C primary-source search excerpts consulted; the PostgreSQL constraints source was consulted earlier in this writing project. Full editorial review remains pending. This entry uses the broad information-management meaning, not biometric identification or statistical model identification.

Previous: Identification in the limit Next: Identifier Domain