ISO/IEC 22989:2022: AI Concepts and Terminology — LinkML Schema

A LinkML schema modelling the artificial-intelligence concepts, terminology, life-cycle stages, stakeholder roles, ecosystem components and application domains defined in ISO/IEC 22989:2022. The schema supplies the controlled vocabulary referenced by sibling lmodel schemas (iso42001 AIMS, iso23894 AI risk management) and supports SSSOM mappings to adjacent AI taxonomies.

URI: https://w3id.org/lmodel/iso22989

Name: iso22989

Classes

Class	Description
AIConceptsCollection	Top-level container aggregating AI systems, models, datasets, lifecycle proce...
NamedEntity	Abstract base class for any addressable entity in the schema, carrying identi...
AbbreviatedTerm	Abbreviation or acronym listed in Clause 4 with its expansion and optional de...
AbbreviationEntry	Record of a single abbreviation listed in Clause 4 of the standard
Action	Action carried out as a result of an AI-system decision (Clause 7
AIApplication	Description of an AI application instance situated in a domain (Clause 10)
AIComponent	Functional component of an AI system, such as a data pipeline, preprocessor, ...
AIConcept	Abstract base for Clause 5 conceptual entities (agent, knowledge, cognition, ...
AIAgent	Entity that perceives its environment and acts upon it to achieve goals (Clau...
CognitiveComputingSystem	System combining AI techniques to emulate human cognitive functions (Clause 5
ConvolutionOperation	Convolution operation as used in convolutional neural networks (Clause 3
HumanMachineTeam	Collaboration arrangement combining one or more humans with one or more AI sy...
IntelligenceAugmentation	Use of AI to enhance the cognitive capabilities of humans rather than replace...
KnowledgeRepresentation	Representation of knowledge usable by an AI system (Clause 5
KnowledgeGraph	Graph-structured knowledge representation, often used for reasoning and retri...
NaturalLanguage	Natural language treated as an object of processing or generation by an AI sy...
Neuron	Computational unit in a neural network combining weighted inputs with a bias ...
SemanticComputingSystem	System whose behaviour is driven by the explicit semantics of its inputs and ...
SoftComputingSystem	System employing soft computing techniques tolerant of imprecision and uncert...
AIEcosystem	Aggregation of the AI systems, data sources, computing resources and stakehol...
AILifecycleProcess	Process or activity associated with a stage of the AI system life cycle (Clau...
AIModel	Trained or rule-based model embedded in an AI system
NeuralNetworkModel	AIModel realised as a neural network (Clause 5
AIStakeholderRole	Stakeholder role enacted by an organisation or individual in relation to an A...
AICustomer	Stakeholder using an AI system or AI-backed service (Clause 5
AIUser	End user of an AI system or AI-backed service (Clause 5
AIPartner	Stakeholder providing supporting services across the AI life cycle (Clause 5
AIAuditor	Partner performing independent audits of AI systems (Clause 5
AIEvaluator	Partner performing evaluations of AI system performance and trustworthiness (...
AISystemIntegrator	Partner integrating AI components into a wider system (Clause 5
DataProvider	Partner supplying datasets used by AI systems (Clause 5
AIProducer	Stakeholder designing, developing or assembling AI systems (Clause 5
ComputationVerifier	Producer role verifying the computational behaviour of an AI system (Clause 5
ModelDesigner	Producer role responsible for designing AI models (Clause 5
ModelImplementer	Producer role responsible for implementing AI models in code (Clause 5
ModelVerifier	Producer role verifying that models meet specified requirements (Clause 5
AIProvider	Stakeholder making an AI system available to customers (Clause 5
AIPlatformProvider	Provider of platform infrastructure on which AI services are operated (Clause...
AIServiceProductProvider	Provider of an AI-enabled service or product to customers (Clause 5
AISubject	Person or group affected by an AI system (Clause 5
DataSubject	Individual whose personal data is processed by an AI system (Clause 5
RelevantAuthority	Regulator or standards-setting body with oversight responsibilities (Clause 5
PolicyMaker	Authority defining policy applicable to AI systems (Clause 5
Regulator	Authority responsible for regulatory oversight of AI systems (Clause 5
AISystem	Engineered system that uses AI techniques to perform tasks delegated to it
AutonomyAssessment	Structured assessment of the autonomy level of an AI system using the criteri...
CatastrophicForgetting	Phenomenon by which a continually-trained model loses previously acquired com...
ComputerVisionFunction	Computer-vision capability provided by an AI system (Clause 9
CyberPhysicalSystem	System that tightly integrates computational and physical components, typical...
DataDrift	Observed change in the statistical distribution of operational data relative ...
DataLabel	Label or annotation attached to one or more data samples (Clause 3
DataProcess	Discrete data-handling process applied to a dataset during AI system developm...
DataSample	Individual data record within a dataset (Clause 3
Dataset	Collection of data items used by an AI system in a training, validation, test...
Decision	Decision produced by an AI system on the basis of one or more predictions (Cl...
EvaluationMetric	Metric used to evaluate AI system or model performance (Clause 7
ExpertSystem	Rule-based system encoding domain expertise (Clause 8
FaultToleranceMechanism	Mechanism enabling an AI system to continue operating correctly in the presen...
GroundTruthRecord	Trusted reference record used to evaluate or train an AI model (Clause 3
Inference	Act of deriving conclusions, predictions or recommendations from a model or k...
InferenceEngine	Component performing inference over a model or knowledge base (Clause 3
InputData	Data presented to an AI system at inference time or during training (Clause 3
InterestedParty	Person or organisation that can affect, be affected by, or perceive itself to...
IoTDevice	Device participating in an Internet-of-Things deployment (Clause 5
IoTSystem	Networked system composed of IoT devices, possibly enhanced with AI capabilit...
NLPComponent	Component of a natural-language-processing pipeline (Clause 9
OECDLifecycleMapping	Informative mapping between an ISO/IEC 22989 life-cycle stage and an OECD lif...
Organization	Organisation that establishes and operates an AI management system; Annex SL ...
Prediction	Prediction produced by an AI model (Clause 7
Recommendation	Recommendation produced by an AI system (Clauses 7
ResourcePool	Pool of computing resources (CPU/GPU/TPU/ASIC/FPGA) available to AI workloads...
RiskItem	Risk associated with an AI system, capturing source, potential event, consequ...
Robot	Embodied agent able to perceive its environment and act in the physical world...
Task	AI task addressed by a model or system (e
Term	Abstract base class for a glossary term defined in Clause 3
AITerm	Term defined in Clause 3
ComputerVisionTerm	Term defined in Clause 3
DataTerm	Term defined in Clause 3
MachineLearningTerm	Term defined in Clause 3
NeuralNetworkTerm	Term defined in Clause 3
NLPTerm	Term defined in Clause 3
TrustworthinessTerm	Term defined in Clause 3
Threshold	Decision threshold applied to a metric, prediction or score (Clause 7
TrustworthinessProperty	Claim about a trustworthiness property of an AI system or model, with evidenc...
VerificationValidationFramework	Verifiability and validatability claim for an AI system characterised accordi...

Slots

Slot	Description
abbreviation_code	Acronym or abbreviation code (Clause 4)
abbreviations	Clause 4 abbreviation records
action_target	Entity or system on which an action is performed (Clause 7
activation_function	Predominant activation function used in the network (Clause 3
actuating_capabilities	Actuating capabilities of the device
affected_dataset	Dataset in which drift was observed
affected_model	Model in which the phenomenon was observed
affected_stakeholders	Stakeholder roles affected by the risk
agent_architecture	Agent architecture realised by the entity (Clause 5
ai_applications	AI application records in the collection
ai_components	AI components that operate within the IoT system
ai_datasets	Datasets referenced by entities in the collection
ai_field	AI sub-field(s) the system draws on
ai_lifecycle_processes	Life-cycle process records in the collection
ai_models	Trained or knowledge-based models in the collection
ai_stakeholder_roles	Stakeholder role records in the collection
ai_system_type	Capability classification of the AI system
ai_systems	AI systems documented in the collection
ai_systems_involved	AI systems participating in the team
algorithm_family	Algorithm family the model belongs to
aliases	Alternative names or synonyms for the term
annotator	Role of the party that produced the label
applicable_biases	Bias categories considered relevant to this property assessment
application_domain	Application domain(s) the AI system targets
applies_to_metric	Name of the metric or score the threshold applies to
approval_criteria	Criteria that must be met for an output or process to be approved
augmented_capability	Cognitive capabilities being augmented
autonomy_criterion_scores	Free-form scores or judgements for autonomy criteria (Clause 5
autonomy_level	Operational autonomy level of the AI system
based_on_predictions	Predictions that supported the decision
big_data_characteristics	Big-data characteristics that the ecosystem exhibits (Clause 8
capacity_units	Capacity expressed in units appropriate to the resource type
catastrophic_forgetting_risk	Estimated risk of catastrophic forgetting on retraining (Clause 5
clause_reference	ISO/IEC 22989:2022 clause identifier (e
cognitive_capabilities	Cognitive capabilities the system provides (e
collection_date_range	Free-text date range over which the data was collected
component_function	Functional view component implemented by an AI component (Clause 7)
components	Constituent components of the AI system
computing_resources	Computing resource types relied upon
confidence	Normalised confidence value associated with an output or claim
confidence_score	Normalised confidence in the property claim
consequence	Consequence to one or more stakeholders if the event occurs
consumed_by	AI system that consumes the input
contact	Contact identifier (e
contains_personal_data	Whether the dataset contains personal or personally identifiable information
controlled_by	AI system that controls the robot
coverage_scope	Scope of failure modes the mechanism covers
cv_task	Type of computer-vision task (Clause 9
cyber_components	Computational components participating in the system
cyber_physical_systems	Cyber-physical system records in the collection
data_collection_method	Method used to collect data (Clause 8
data_modality	Modalities present in the dataset
data_processes	Data-handling processes applied within the AI system (Clause 5
data_processes_applied	Data-handling processes applied to the dataset (Clause 5
data_provenance	Provenance statement for the dataset (origin, collection method, licensing)
data_quality_notes	Notes on data quality, completeness or representativeness
data_source_type	Classification of a data source (Clause 8
data_sources	Identifiers or descriptions of data sources feeding the ecosystem
data_version	Version identifier of a dataset snapshot
dataset_role	Role the dataset plays in the ML workflow
datasets	Datasets used by, or produced by, the AI system
decision_outcome	Chosen course of action resulting from a decision (Clause 7
decision_policy	Policy or rule used to translate predictions into a decision
depends_on	Other components this component depends on at runtime
description	Free-text description of the entity (paraphrased; verbatim ISO text excluded)
detected_at	Timestamp or interval at which drift was detected
device_role	Role played by a device in an IoT or cyber-physical system (Clause 5
devices	Devices that make up the IoT system
drift_type	Type of drift (covariate, label, concept)
ecosystem_components	Free-text or CURIE references to ecosystem components
edge_count	Approximate number of edges in a graph-structured artefact
embodiment	Free-text description of the physical embodiment
end_date	Date the process completed
engineering_approach	Non-learning engineering approach used (for symbolic/knowledge-based models)
executed_by	Stakeholder role that executed the process
execution_status	Execution status of an action or process
expansion	Expanded form of an abbreviation (Clause 4)
expert_systems	Expert-system records in the collection
feature_count	Number of features per record in a dataset
functional_components	Functional components exhibited by the system (Clause 7)
goal_set	Goals the agent is configured to pursue
ground_truth_available	Whether trusted ground-truth labels are available (Clause 3
ground_truth_value	Trusted reference value (Clause 3
hosting_system	AI system that provides the application
human_roles	Roles played by humans in the team
hyperparameters	Free-form record of model hyperparameter settings
id	Unique CURIE or URI identifying the entity
inference_engine	Inference engine used by the expert system
inference_latency_ms	Typical end-to-end inference latency in milliseconds
inference_strategy	Inference strategy used (e
input_arity	Number of inputs combined by the neuron
input_dataset	Dataset consumed by the process
input_modalities	Modalities of input accepted by a task or component
intended_purpose	Stated intended purpose of the AI system
iot_integration	IoT/CPS system this AI system is integrated with, if any (Clause 5
iot_subsystem	IoT subsystem the CPS relies on, if any
iot_systems	IoT system records in the collection
iso_stage	ISO/IEC 22989 life-cycle stage
jurisdictional_issues	Jurisdictional issues considered in scope for the application
kernel_size	Spatial dimensions of the convolution kernel
knowledge_graphs	Knowledge-graph records in the collection
knowledge_type	Type of knowledge captured (Clause 3
label_type	Type of target label associated with a dataset, sample or annotation (Clause ...
label_value	Concrete label value
language_code	BCP-47 language tag
lifecycle_stage	Current life-cycle stage of the AI system
likelihood	Estimated likelihood of the event (0
mapping_notes	Free-text notes on the mapping relationship
measurement_method	How the property was assessed or measured
mechanism_type	Type of mechanism (redundancy, graceful degradation, failover, etc
metric_name	Name of the metric (e
metric_unit	Unit of the metric, when applicable
metric_value	Observed numeric value of the metric
mitigation_strategy	Strategy applied to mitigate the phenomenon
modality	Modality of the input data
model_compression_applied	Whether the model has had compression or distillation applied for deployment ...
model_paradigm	Machine-learning paradigm under which the model was trained
model_version	Version identifier of the trained model artefact
models	Trained or knowledge-based models embedded in the AI system
name	Human-readable label for the entity
neural_network_architecture	Neural-network architecture, when algorithm_family is neural_network (Clauses...
nlp_component_type	Type of NLP pipeline component (Clause 9
node_count	Approximate number of nodes in a graph-structured artefact
number_of_layers	Number of layers in the network
number_of_parameters	Approximate number of trainable parameters
oecd_stage	Corresponding OECD life-cycle stage
ontology_reference	Ontologies referenced by a knowledge artefact
organization_name	Name of the organisation acting in the stakeholder role
output_dataset	Dataset produced by the process
output_label_type	Type of output label produced by a task, when applicable
over_model	Model over which the inference was performed
padding	Padding mode (e
parameter_count	Approximate count of trainable model parameters
parameters	Free-form parameters configuring the process
performance_metric	Metrics used to evaluate performance
performed_by	Engine that performed the inference
physical_processes	Physical processes the system monitors or controls
potential_event	Potential event whose occurrence would realise the risk
predicted_value	Serialised representation of a predicted value (Clause 7
preferred_label	Preferred natural-language label
process_inputs	Inputs consumed by the process
process_outputs	Outputs produced by the process
process_stage	Life-cycle stage the process belongs to
process_sub_type	Free-text refinement of the process within its life-cycle stage (e
process_type	Type of data-handling process performed
produced_by	Model that produced the prediction
produced_output	Serialised representation of the inference output
property_evidence	References to evidence supporting the property claim
provenance_statement	Provenance description for an artefact
recommendation_outcome_type	High-level outcome type of a recommendation (Clauses 7
recommended_items	Items recommended by the system, serialised as strings
record_count	Number of records or examples in the dataset
reference_dataset	Dataset against which the metric was computed
representation_form	Concrete form of representation (rules, frames, ontology, graph, vectors)
resource_type	Category of computing resource (Clause 8
responsibilities	Free-text statements of responsibility
responsible_role	Stakeholder role responsible for executing the process
risk_items	Risks identified or addressed by a process or assessment
risk_source	Source from which the risk originates
rule_count	Approximate count of rules in a rule-based knowledge base
sample_label	Label or target value associated with the sample
sample_payload	Serialised representation of the sample contents
script	ISO 15924 script code, when relevant
see_also_uri	Pointers to related external resources or term records
semantic_model	Reference to the semantic model or ontology used
sensing_capabilities	Sensing capabilities of the device
severity	Qualitative severity assessment of the consequence
societal_impacts	Societal impact categories considered in scope
soft_computing_techniques	Soft-computing techniques the system employs
stakeholder_role_type	Canonical stakeholder role type as defined in Clause 5
stakeholders	Stakeholder roles associated with the AI system
start_date	Date the process started
stride	Stride applied when sliding the kernel over the input
supports_continuous_learning	Whether a model supports continuous or online learning (Clause 3
symbolic_approach	Predominant symbolic vs subsymbolic approach used by the system
system_characteristics	Distinguishing characteristics from Clause 5
task_allocation	Free-text description of how tasks are allocated between humans and AI
task_categories	Task categories the AI system addresses
task_category	Category of AI task being addressed
tasks	Task records in the collection
test_dataset	Dataset used to estimate generalisation performance
threshold_policy	Policy describing how the threshold is interpreted
threshold_value	Numeric threshold value
trained_on	Date or version reference for when the model was last trained
training_dataset	Dataset used to fit the model
training_duration	Wall-clock duration of model training, expressed as an ISO 8601 duration
training_phenomena	Training-time phenomena observed for the model (Clause 3
triggered_by	Decision that triggered the action
trustworthiness_properties	Trustworthiness properties claimed for the AI system
trustworthiness_property_type	Which trustworthiness property is being characterised
trustworthiness_records	Trustworthiness property claims documented in the collection
uses_model	Model the inference engine evaluates
validation_dataset	Dataset used for hyperparameter tuning and model selection
validation_methods	Validation methods applied or applicable to the system
validation_strategy	Strategy used to estimate generalisation (Clause 5
verification_methods	Verification methods applied or applicable to the system
verification_validation_level	Verifiability / validatability claim (Clause 5

Enumerations

Enumeration	Description
AbbreviationCode	Acronyms and abbreviations listed in Clause 4
ActivationFunctionType	Common activation functions used in neural networks (Clause 3
AgentArchitectureType	Agent architectures discussed in Clause 5
AIApplicationDomain	Example AI application domains presented in Clause 10
AIField	Sub-fields of AI referenced in Clause 9
AIFunctionalComponent	Functional building blocks of an AI system as introduced in Clause 7
AILifecycleStage	AI system life-cycle stages identified in Clause 6
AIStakeholderRoleType	AI stakeholder roles enumerated in Clause 5
AISystemCharacteristic	Distinguishing characteristics of AI systems summarised in Clause 5
AISystemType	High-level capability classification of an AI system, from narrow (single-tas...
AutonomyCriterion	Criteria contributing to the assessment of autonomy in Clause 5
AutonomyLevel	Degree of system autonomy as discussed in Clause 5
BiasType	Categories of bias relevant to AI systems as discussed in Clause 5
BigDataCharacteristic	Characteristics commonly used to describe big data sources in Clause 8
ComputerVisionTask	Computer-vision tasks drawn from Clauses 3
ComputingResourceType	Categories of computing resource used by AI systems, drawn from Clauses 8
DataCollectionMethod	Methods of data collection enumerated in Clause 8
DataLabelType	Categories of target label produced or consumed by ML workflows
DataModality	Modalities of input data handled by AI systems, drawn from the data, NLP and ...
DataProcessType	Data-handling processes enumerated in Clause 5
DatasetRole	Role a dataset plays in a machine-learning workflow, drawn from Clauses 5
DataSourceType	Classification of data sources discussed in Clause 8
EngineeringApproach	Non-learning engineering approaches contributing to AI, from Clause 8
ExecutionStatus	Execution status values for actions and processes
IoTDeviceRole	Roles played by devices in IoT and cyber-physical systems (Clause 5
JurisdictionalIssueType	Categories of jurisdictional issue surfaced in Clause 5
KnowledgeType	Types of knowledge distinguished in Clause 3
MachineLearningParadigm	Top-level machine-learning paradigms enumerated in Clause 5
MLAlgorithmFamily	Example machine-learning algorithm families enumerated in Clause 5
NeuralNetworkArchitecture	Architectural families of neural networks enumerated across Clause 3
NeuralNetworkPhenomenon	Training-time phenomena that affect neural-network learning, drawn from Claus...
NeuroSymbolicApproach	Sub-categorisation of hybrid neuro-symbolic approaches mentioned in\n Cl...
NLPComponentType	Components of a natural-language-processing pipeline as enumerated in Clauses...
OECDLifecycleStage	OECD AI system life-cycle stages used in the informative mapping of Annex A
RecommendationOutcomeType	High-level categorisation of recommendations produced by an AI system\n ...
SocietalImpactCategory	Categories of societal impact discussed in Clause 5
SoftComputingTechnique	Techniques grouped under soft computing in Clause 5
SymbolicApproach	The symbolic vs subsymbolic axis used in Clause 5
TaskCategory	Categories of AI task addressed by AI systems, derived from the\n machin...
TrustworthinessPropertyType	Properties contributing to AI trustworthiness, enumerated in Clause 5
ValidationStrategy	Strategies for partitioning data and assessing generalisation, drawn from Cla...
VerificationValidationLevel	Levels of verifiability and validatability used to characterise an AI system ...

Types

Type	Description
Boolean	A binary (true or false) value
ConfidenceScore	Numeric confidence score expressed as a value in the closed interval [0
Curie	a compact URI
Date	a date (year, month and day) in an idealized calendar
DateOrDatetime	Either a date or a datetime
Datetime	The combination of a date and time
Decimal	A real number with arbitrary precision that conforms to the xsd:decimal speci...
Double	A real number that conforms to the xsd:double specification
DurationType	ISO 8601 duration value such as P1Y, P30D or PT4H
Float	A real number that conforms to the xsd:float specification
Integer	An integer
Jsonpath	A string encoding a JSON Path
Jsonpointer	A string encoding a JSON Pointer
Ncname	Prefix part of CURIE
Nodeidentifier	A URI, CURIE or BNODE that represents a node in a model
Objectidentifier	A URI or CURIE that represents an object in the model
Sparqlpath	A string encoding a SPARQL Property Path
String	A character string
Time	A time object represents a (local) time of day, independent of any particular...
Uri	a complete URI
Uriorcurie	a URI or a CURIE

Subsets

Subset	Description
AiApplicationsDomain	Application domains presented in Clause 10 (fraud detection, automated vehicl...
AiConcepts	Conceptual elements introduced in Clause 5 (AI concepts), including agent, kn...
AiEcosystem	Ecosystem-level classes in Clause 8 (AI systems, AI functions, ML, engineerin...
AiFields	Sub-fields of AI introduced in Clause 9 (computer vision, NLP, data mining, p...
AiFunctionalView	Classes representing the functional view of an AI system in Clause 7 (data an...
AiLifecycle	Classes and enums representing the AI system life-cycle model and its stages ...
AiStakeholders	AI stakeholder roles enumerated in Clause 5
Terminology	Classes and enums representing terms and definitions in Clause 3 of ISO/IEC 2...
Trustworthiness	Trustworthiness properties of AI systems as listed in Clause 5