# The seven ISO 42102 defined terms

Home / Defined terms

# The seven defined terms

ISO/IEC DIS 42102 defines seven terms of its own and imports the rest from ISO/IEC 22989:2022. Every one of the seven has a page here with the meaning in our own words, what it settles, the evidence to keep and a test to apply to your own documents.

The seven are the load-bearing words of the framework: algorithm, AI system capabilities, capability, conversational learning, emergent behaviour, matrix and hybrid AI. Capability is the unit on one axis, matrix is the shape of the whole, and hybrid AI plus conversational learning mark where one method label stops being enough.

These definitions are public and can be read before the full draft is available. They are also the part most likely to be used loosely in an organisation, so each page includes a test that catches the common misuse rather than only restating the meaning.

Clause 3 of the public preview
CodeTermMeaning in our words
3.1algorithmA finite, ordered set of instructions that accomplishes a task or solves a problem. The draft imports the idea from the older data processing vocabulary and adapts it to the AI context.
3.2AI system capabilitiesThe collective term for what an AI system can do: the abilities and functionalities attributed to it. The entry is a synonym pointer, so the operative meaning sits in the entry for capability.
3.3capabilityThe ability of a system to carry out a task. It is a property of the system, stated as what it can do, not as how it does it.
3.4conversational learningA hybrid method in which a system improves through interactive dialogue with users or other agents, taking in new knowledge during the conversation rather than only through a separate training cycle.
3.5emergent behaviourObservable behaviour that appears when one or more AI systems coordinate several capabilities of their components across more than one knowledge domain. The note names factual, conceptual, procedural and metacognitive domains.
3.6matrixA two-dimensional picture that sets the methods of AI systems against their capabilities. It is the shape of the description, not a score and not a maturity model.
3.7hybrid AIA method that combines mechanisms from different families: learning approaches, knowledge forms, neural designs, symbolic reasoning or statistical models.

The imported vocabulary still governs. The draft adopts ISO/IEC 22989:2022 for the terms it does not define. A firm that uses this framework without that vocabulary is using half a language.

Last updated: 2026-10-22

Source: https://iso-42102-revision-checker.theartofservice.com/glossary/