oSIST prEN ISO/IEC 42102:2026

• Abstract:

  This document provides a framework of descriptors to support the consistent characterization of AI system methods and capabilities. The framework helps AI stakeholders describe AI systems and have a common understanding of them. This document applies to all types of organizations involved in any of the lifecycle stages of AI systems as well as to any AI stakeholder roles

• Status: Not Published
• Public Enquiry End Date: 07-Sep-2026
• ICS: 35.020 - Information technology (IT) in general
• ICS: 35.240.01 - Application of information technology in general
• Technical Committee: UMI - Artificial intelligence
• Current Stage: 4020 - Public enquire (PE) (Adopted Project)
• Start Date: 26-Jun-2026
• Due Date: 13-Nov-2026
• Completion Date: 20-Aug-2026
• Ref Project: prEN ISO/IEC 42102 - Information technology - Artificial intelligence - Framework for characterizing AI system methods and capabilities (ISO/IEC DIS 42102:2026)

Overview

oSIST prEN ISO/IEC 42102:2026:2026 - Information technology – Artificial intelligence – Framework for characterizing AI system methods and capabilities is a forthcoming international standard developed by CEN under the ISO/IEC JTC 1/SC 42 technical committee. This document provides a structured framework for consistently describing AI system methods and capabilities, aimed at facilitating communication, understanding, and effective management of AI technologies across organizations and industries.

The standard delivers a common language and a comprehensive set of descriptors, enabling stakeholders-such as developers, regulators, purchasers, and users-to clearly characterize and assess AI systems throughout all stages of their lifecycle. This framework is essential for aligning stakeholder perspectives, supporting risk management, and informing decision-making in the evolving field of artificial intelligence.

Key Topics

• Framework of Descriptors: A structured approach for defining both the methods (algorithms, procedures) and capabilities (functional abilities) of AI systems.
• Dimensions of Characterization: The standard distinguishes between AI methods (e.g., learning strategies, reasoning algorithms, representational forms) and AI capabilities (e.g., sensory processing, knowledge management, action execution).
• Consistency in Description: By using standardized descriptors, organizations can ensure that AI systems are described uniformly, which supports transparency and interoperability.
• Stakeholder Communication: The framework supports productive dialogue between technical and non-technical stakeholders, including industry, academia, public administration, and civil society.
• Applicability: The framework applies to all types of organizations and roles involved in any lifecycle stage of AI systems, fostering a unified foundation for AI management and governance.

Applications

The practical value of oSIST prEN ISO/IEC 42102:2026 lies in its broad applicability and its role in enhancing the trustworthy and responsible use of AI. Key applications include:

• AI Project Planning and Assessment: Organizations can use the framework to plan, document, and evaluate AI projects by accurately characterizing system methods and capabilities.
• Conformity Assessments & Compliance: The descriptors facilitate assessments related to conformity, regulatory compliance, and incident management.
• Risk Management: By clarifying methods and capabilities, stakeholders can better identify, communicate, and mitigate potential risks associated with AI systems.
• Benchmarking and Evaluation: The standard supports the comparison of different AI solutions, making it easier to select appropriate technologies for specific needs.
• Policy and Regulation: Policymakers and regulators benefit from a structured vocabulary when drafting or enforcing AI-related legislation.
• Best Practices and Knowledge Sharing: The common framework promotes efficient knowledge transfer and adoption of best practices across organizations, fostering a culture of transparency and continuous improvement in AI.

Related Standards

oSIST prEN ISO/IEC 42102:2026 is closely aligned with other international standards to promote a cohesive approach to AI characterization:

• ISO/IEC 22989:2022 – Information technology – Artificial intelligence – Concepts and terminology. This standard establishes the foundational terminology referenced within oSIST prEN ISO/IEC 42102:2026.
• ISO/IEC 2382:2015 – Information technology – Vocabulary. Provides key definitions used in the framework.
• Other AI Management and Safety Standards: oSIST prEN ISO/IEC 42102:2026 complements standards that focus on AI lifecycle management, safety, risk assessment, and system performance.

---

In summary, oSIST prEN ISO/IEC 42102:2026:2026 provides a critical foundation for consistent, transparent, and effective characterization of AI systems. Its adoption will support robust governance, foster cross-sector collaboration, and drive the responsible development and deployment of artificial intelligence technologies worldwide.

SLOVENSKI STANDARD
01-september-2026
Informacijska tehnologija - Umetna inteligenca - Okvir za opredelitev metod in
zmožnosti sistemov umetne inteligence (ISO/IEC DIS 42102:2026)
Information technology - Artificial intelligence - Framework for characterizing AI system
methods and capabilities (ISO/IEC DIS 42102:2026)
Informationstechnik - Künstliche Intelligenz - Taxonomie der Methoden und Fähigkeiten
von KI-Systemen (ISO/IEC DIS 42102:2026)
Technologies de l'information - Intelligence artificielle - Cadre de caractérisation des
méthodes et capacités des systèmes d'IA (ISO/IEC DIS 42102:2026)
Ta slovenski standard je istoveten z: prEN ISO/IEC 42102
ICS:
35.020 Informacijska tehnika in Information technology (IT) in
tehnologija na splošno general
2003-01.Slovenski inštitut za standardizacijo. Razmnoževanje celote ali delov tega standarda ni dovoljeno.

DRAFT
International
Standard
ISO/IEC DIS 42102
ISO/IEC JTC 1/SC 42
Information technology — Artificial
Secretariat: ANSI
intelligence — Framework for
Voting begins on:
characterizing AI system methods
2026-06-15
and capabilities
Voting terminates on:
ICS: 35.240.01
2026-09-07
THIS DOCUMENT IS A DRAFT CIRCULATED
FOR COMMENTS AND APPROVAL. IT
IS THEREFORE SUBJECT TO CHANGE
AND MAY NOT BE REFERRED TO AS AN
INTERNATIONAL STANDARD UNTIL
PUBLISHED AS SUCH.
This document is circulated as received from the committee secretariat.
IN ADDITION TO THEIR EVALUATION AS
BEING ACCEPTABLE FOR INDUSTRIAL,
TECHNOLOGICAL, COMMERCIAL AND
USER PURPOSES, DRAFT INTERNATIONAL
STANDARDS MAY ON OCCASION HAVE TO
ISO/CEN PARALLEL PROCESSING
BE CONSIDERED IN THE LIGHT OF THEIR
POTENTIAL TO BECOME STANDARDS TO
WHICH REFERENCE MAY BE MADE IN
NATIONAL REGULATIONS.
RECIPIENTS OF THIS DRAFT ARE INVITED
TO SUBMIT, WITH THEIR COMMENTS,
NOTIFICATION OF ANY RELEVANT PATENT
RIGHTS OF WHICH THEY ARE AWARE AND TO
PROVIDE SUPPORTING DOCUMENTATION.
Reference number
© ISO/IEC 2026
ISO/IEC DIS 42102:2026(en)
DRAFT
ISO/IEC DIS 42102:2026(en)
International
Standard
ISO/IEC DIS 42102
ISO/IEC JTC 1/SC 42
Information technology — Artificial
Secretariat: ANSI
intelligence — Framework for
Voting begins on:
characterizing AI system methods
and capabilities
Voting terminates on:
ICS: 35.240.01
THIS DOCUMENT IS A DRAFT CIRCULATED
FOR COMMENTS AND APPROVAL. IT
IS THEREFORE SUBJECT TO CHANGE
AND MAY NOT BE REFERRED TO AS AN
INTERNATIONAL STANDARD UNTIL
PUBLISHED AS SUCH.
This document is circulated as received from the committee secretariat.
IN ADDITION TO THEIR EVALUATION AS
BEING ACCEPTABLE FOR INDUSTRIAL,
© ISO/IEC 2026
TECHNOLOGICAL, COMMERCIAL AND
USER PURPOSES, DRAFT INTERNATIONAL
All rights reserved. Unless otherwise specified, or required in the context of its implementation, no part of this publication may
STANDARDS MAY ON OCCASION HAVE TO
ISO/CEN PARALLEL PROCESSING
be reproduced or utilized otherwise in any form or by any means, electronic or mechanical, including photocopying, or posting on
BE CONSIDERED IN THE LIGHT OF THEIR
the internet or an intranet, without prior written permission. Permission can be requested from either ISO at the address below
POTENTIAL TO BECOME STANDARDS TO
WHICH REFERENCE MAY BE MADE IN
or ISO’s member body in the country of the requester.
NATIONAL REGULATIONS.
ISO copyright office
RECIPIENTS OF THIS DRAFT ARE INVITED
CP 401 • Ch. de Blandonnet 8
TO SUBMIT, WITH THEIR COMMENTS,
CH-1214 Vernier, Geneva
NOTIFICATION OF ANY RELEVANT PATENT
Phone: [phone]
RIGHTS OF WHICH THEY ARE AWARE AND TO
PROVIDE SUPPORTING DOCUMENTATION.
Email: [e-mail]
Website: www.iso.org
Published in Switzerland Reference number
© ISO/IEC 2026
ISO/IEC DIS 42102:2026(en)
© ISO/IEC 2026 – All rights reserved
ii
ISO/IEC DIS 42102:2026(en)
Contents Page
Foreword .v
Introduction .vi
1 Scope . 1
2 Normative references . 1
3 Terms and definitions . 1
4 Principles of AI system characterization based on methods and capabilities . 2
4.1 General .2
4.2 Definition of principles . .2
4.3 Intended application of principles.3
4.3.1 General .3
4.3.2 Interweaving of methods with components, tasks, algorithms and models .3
4.4 Understanding AI methods and AI system capabilities .3
4.5 Application to AI systems .4
5 AI system methods dimension . 4
5.1 General .4
5.2 Traditional AI .5
5.2.1 General .5
5.2.2 Search .6
5.2.3 Optimization .6
5.2.4 Planning and plan recognition .6
5.2.5 Decision-making .7
5.3 Symbolic AI .7
5.3.1 General .7
5.3.2 Knowledge representation .8
5.3.3 Logical reasoning .8
5.3.4 Probabilistic reasoning .8
5.3.5 Non-probabilistic reasoning .8
5.4 Machine learning .9
5.4.1 General .9
5.4.2 Machine learning model characterization according to training and inference .9
5.4.3 Reinforcement learning .10
5.4.4 Supervised learning .11
5.4.5 Unsupervised learning.11
5.4.6 Semi-supervised learning .11
5.4.7 Self-supervised learning .11
5.5 Hybrid AI .11
5.5.1 General .11
5.5.2 Hybrid neuronal systems . 12
5.5.3 Learning with knowledge . 12
5.5.4 Conversational learning . 12
6 AI system capabilities dimension .12
6.1 General . 12
6.2 Sensory input processing . 13
6.2.1 General . 13
6.2.2 External sensory input processing . 13
6.2.3 Internal sensory processing . .14
6.3 Knowledge processing .14
6.3.1 General .14
6.3.2 Factual knowledge .14
6.3.3 Conceptual knowledge . 15
6.3.4 Procedural knowledge . 15
6.3.5 Metacognitive knowledge . 15
6.4 Action control processing .16

© ISO/IEC 2026 – All rights reserved
iii
ISO/IEC DIS 42102:2026(en)
6.4.1 General .16
6.4.2 Physical realization and informational control .16
6.5 Communication processing .16
6.5.1 General .16
6.5.2 Communication processing without feedback .16
6.5.3 Communication processing with feedback .17
6.6 Emergent behaviour .17
6.6.1 General .17
6.6.2 Analysis .17
6.6.3 Optimization .17
6.6.4 Decision-making .18
6.6.5 Search and problem-solving .18
6.6.6 Planning and plan-recognition .18
6.6.7 Creativity .18
6.6.8 Learning and adjustment .19
6.6.9 Explanation .19
6.6.10 Diagnosis .19
6.6.11 Multimodality .19
6.6.12 Generative AI behaviour . 20
7 Framework for characterizing AI system methods and capabilities .20
7.1 General . 20
7.2 Relation between AI tasks, AI components, AI methods and AI capabilities .21
7.3 Characterization framework of AI system methods and capabilities . 22
7.4 Example of using the framework: humanoid soccer robot . 23
7.5 Use of framework alongside further standardization documents .24
Annex A (informative) Dimensions of observation .26
Annex B (informative) Application of this document .28
Annex C (informative) Properties of machine learning methods .35
Annex D (informative) AI paradigms .37
Bibliography .39

© ISO/IEC 2026 – All rights reserved
iv
ISO/IEC DIS 42102:2026(en)
Foreword
ISO (the International Organization for Standardization) and IEC (the International Electrotechnical
Commission) form the specialized system for worldwide standardization. National bodies that are
members of ISO or IEC participate in the development of International Standards through technical
committees established by the respective organization to deal with particular fields of technical activity.
ISO and IEC technical committees collaborate in fields of mutual interest. Other international organizations,
governmental and non-governmental, in liaison with ISO and IEC, also take part in the work.
The procedures used to develop this document and those intended for its further maintenance are described
in the ISO/IEC Directives, Part 1. In particular, the different approval criteria needed for the different types
of document should be noted. This document was drafted in accordance with the editorial rules of the ISO/
IEC Directives, Part 2 (see www.iso.org/directives or www.iec.ch/members_experts/refdocs).
Attention is drawn to the possibility that some of the elements of this document may be the subject of patent
rights. ISO and IEC shall not be held responsible for identifying any or all such patent rights. Details of any
patent rights identified during the development of the document will be in the Introduction and/or on the
ISO list of patent declarations received (see www.iso.org/patents) or the IEC list of patent declarations
received (see https://patents.iec.ch).
Any trade name used in this document is information given for the convenience of users and does not
constitute an endorsement.
For an explanation of the voluntary nature of standards, the meaning of ISO specific terms and expressions
related to conformity assessment, as well as information about ISO's adherence to the World Trade
Organization (WTO) principles in the Technical Barriers to Trade (TBT) see www.iso.org/iso/foreword.html.
In the IEC, see www.iec.ch/understanding-standards.
This document was prepared by Joint Technical Committee ISO/IEC JTC 1, Information technology,
Subcommittee SC 42, Artificial intelligence.
Any feedback or questions on this document should be directed to the user’s national standards
body. A complete listing of these bodies can be found at www.iso.org/members.html and
www.iec.ch/national-committees.

© ISO/IEC 2026 – All rights reserved
v
ISO/IEC DIS 42102:2026(en)
Introduction
Advances in artificial intelligence (AI) have revolutionized various industries and transformed the way
humans master technology. As AI continues to permeate our daily lives, it is crucial to establish standardized
ways of describing AI systems that enable a shared understanding of these systems, including to better
evaluate value and risk.
This document provides a framework for characterizing AI system methods (procedures, algorithms) and
capabilities (abilities, functionalities). The framework can be used by stakeholders to characterize AI system
methods and capabilities. The aim is to help relevant stakeholders to understand methods and capabilities
used in an AI system. AI systems can consist of several components, each of which suggests one or more
methods. In this context, the framework helps distinguish between different types of AI methods and AI
system capabilities to create a common language for analysing, comparing and communicating about the
diverse landscape of AI systems from a scientific perspective. According to this document, AI methods do
not restrict to machine learning algorithms but also encompass a wide range of approaches, like heuristics,
optimization algorithms, knowledge representation, reasoning and hybrid algorithms. Using these diverse
methods, AI systems function not solely based on data-driven learning. By providing such a framework, this
document aims to simplify the management of AI system’s life cycle stages. By promoting the responsible
and beneficial use of AI systems, this document aims to enhance transparency, foster trustworthiness of AI
systems.
This document provides a common framework of descriptors that assist with communication about AI
systems’ technical characteristics among representatives from industry, science, public administration
and civil society. The framework can be used descriptively and instructively by organizations, researchers,
policymakers, practitioners involved in the development, deployment and regulation of AI systems as well as
AI ecosystems. For example, the descriptors can help recipients draw inferences about matters such as what
applications a system can be suitable for, what processes can be followed during system construction, what
would be necessary in order to make changes to the system or what aspects of the system’s behaviour can be
easy or difficult for humans to form mental models of. Containing a matrix of methods and capabilities, this
document demonstrates that AI systems can possess multiple AI methods and theoretically switch between
them while executing an AI task. The matrix can be used to separate elements of complex systems into
method and capability pairs, helping to understand the correct approach for conformity assessment as well
as incident management and as a tool for communicating compliance-related issues among stakeholders.
The characterization can also be used to distinguish between AI methods in safety-related International
Standards. In the course of using the framework, applicants of the framework can gain a deeper
understanding of AI systems in order to standardize planning and decision processes related to stages of the
AI system life cycle. For instance, the framework can help to better understand the AI system’s methods and
capabilities and to present various AI system characteristics in a more informative way, e.g. conformance,
metrics, risks, technical performance, sustainability, autonomy, quality attributes as well as strengths
and weaknesses. In addition, the framework can be used to identify opportunities for further advances in
research of AI methods and capabilities.
This document does not cover AI systems that are intentionally designed or put in operation to perform
avoidable harmful acts. For instance, harmful acts can include illegal or unethical nudging, surveillance,
sorting, mental as well as psychological harm with regard to people, as well as malicious interference with
data and finances.
In this document, the following structure depicts the methodology of AI system methods and capabilities:
— Clause 4 outlines the framework for characterizing AI system methods and capabilities;
— Clause 5 presents descriptors for AI system methods;
— Clause 6 presents descriptors for AI system capabilities;
— Clause 7 provides framework for characterization of AI system methods and capabilities;
— Informative annexes provide explanations and examples of how the framework can be applied.

© ISO/IEC 2026 – All rights reserved
vi
DRAFT International Standard ISO/IEC DIS 42102:2026(en)
Information technology — Artificial intelligence —
Framework for characterizing AI system methods and
capabilities
1 Scope
This document provides a framework of descriptors to support the consistent characterization of AI system
methods and capabilities. The framework helps AI stakeholders to describe AI systems and have a common
understanding of them.
This document applies to all types of organizations involved in any of the life cycle stages of AI systems as
well as to any AI stakeholder roles.
2 Normative references
The following documents are referred to in the text in such a way that some or all of their content constitutes
requirements of this document. For dated references, only the edition cited applies. For undated references,
the latest edition of the referenced document (including any amendments) applies.
ISO/IEC 22989:2022, Information technology — Artificial intelligence — Artificial intelligence concepts and
terminology
3 Terms and definitions
For the purposes of this document, the terms and definitions given in ISO/IEC 22989:2022 and the following
apply.
ISO and IEC maintain terminology databases for use in standardization at the following addresses:
— ISO Online browsing platform: available at https:// www .iso .org/ obp
— IEC Electropedia: available at https:// www .electropedia .org/
3.1
algorithm
finite sequence of well-defined instructions for accomplishing a task or for solving a problem
[SOURCE: ISO 2382:2015(en), 2121376, modified]
3.2
AI system capabilities
AI capabilities of AI systems
3.3
capability
ability of an AI system to perform a task
3.4
conversational learning
hybrid AI method that improves the function of an AI system through interactive dialog-based feedback
with users or other agents, actively incorporating new knowledge during conversations

© ISO/IEC 2026 – All rights reserved
ISO/IEC DIS 42102:2026(en)
3.5
emergent behaviour
observable system behaviour that arises when one or more AI systems dynamically coordinate multiple
capabilities of their components across several knowledge processing domains
Note 1 to entry: Knowledge processing domains include the factual, conceptual, procedural and metacognitive
domains.
3.6
matrix
two-dimensional concept depicting the methods and capabilities of AI systems
3.7
hybrid AI
merges the mechanisms of different AI methods, integrating paradigms such as learning strategies,
representational forms, neural architectures, symbolic reasoning and statistical models
4 Principles of AI system characterization based on methods and capabilities
4.1 General
The framework categorizes AI component properties into the dimensions of AI methods and AI component
[1] [2] [3] [4]
capabilities. , , , After providing a definition of the individual principles for describing AI systems
using methods and capabilities, the intended implementation of those principles is described. Different
perspectives on observing AI systems are represented in Annex A. Clause 7.4 and Annex B depict how the
framework can be used to describe and understand AI systems.
4.2 Definition of principles
The dimensions in this document have been elaborated upon based on the following principles:
a) Underpinning common communication needs: The dimensions and the descriptors within them
are selected to reflect aspects of the workings of an AI system. AI actors, such as those who develop,
purchase or oversee AI systems, can communicate about these dimensions and descriptors with each
other (e.g. to facilitate inferences about possibilities and limitations).
b) Consistent coverage: The dimensions are meant to provide for description of a wide range of AI systems,
including state-of-the-art systems. This includes allowing for accurate description of existing systems
while still providing flexibility to appropriately classify novel methods or capabilities.
c) Clear distinctions between dimensions: The dimensions represent aspects of AI systems that can
vary somewhat independently. Nonetheless, there can be strong correlations between dimensions.
For example, modern systems that process image inputs using the capabilities dimension are usually
trained using some form of unsupervised or self-supervised learning from the methods dimension. AI
methods refer to the algorithms and procedures by which an AI system performs specific functions.
These methods can be embedded at the system or component level and vary in architecture, ranging
from single-method systems to those combining multiple methods.
d) High-level categories with room for multiple levels of detail: An AI system or AI component can be
described along a given dimension at varying levels of detail. For example, in some communications
about AI systems, it can suffice to specify that a model was built using machine learning. For others it
can be necessary to specify that it was built using stochastic gradient descent. This document specifies
only high-level descriptors within each dimension, but offers examples of more granular distinctions
that can be drawn. It leaves room for adding such detail in glosses and for other documents to elaborate
further on distinctions within a dimension or entirely different ways of systematically describing or
categorizing AI systems.
© ISO/IEC 2026 – All rights reserved
ISO/IEC DIS 42102:2026(en)
4.3 Intended application of principles
4.3.1 General
Based on its principles, the framework provides a structured set of descriptors in terms of which AI systems
can be partially described and makes it easier to differentiate between different types of AI systems. It
encompasses AI methods, which represent the procedures and algorithms employed by AI systems and
AI system capabilities, which describe the specific abilities and functionalities of AI systems. Within the
framework, both AI system methods and AI system capabilities are grouped into subsets with varying
levels of detail. The selection of specific subsets with a particular level of detail depends on the intended
application scenario of the framework. The distinction between methods and capabilities assists in precise
descriptions of AI systems, highlighting their strengths, limitations and implications.
4.3.2 Interweaving of methods with components, tasks, algorithms and models
The following statements reflect the relationships between AI methods, capabilities, models, tasks, systems
and components:
— An AI system is composed of one or more components, at least one of which is an AI component. An AI
component can be a standalone AI system;
— An AI system can contain multiple AI components that interact with each other or non-AI components of
the overall system;
— An AI task is characterized by the input-to-output transformation that an AI system or component is
expected to perform;
— An AI component is a functional component of an AI system. Typically, an AI component is built to
execute one AI task, though the task can be a very general one, such as generating images from textual
descriptions;
— An AI component takes data as input and produces data as output;
— The capabilities dimension is defined in terms of the task or tasks performed by an AI component;
— An AI model is a computational representation that enables an AI method to execute an AI task;
NOTE 1 An AI model is a computational representation that helps AI systems to execute one or more classes of
AI tasks.
— An AI method or AI model method reflects a mathematical algorithm which computes input data to
execute an AI task.
NOTE 2 An AI method is a class of AI models, a class of algorithms used by AI systems often to operate on AI
models or a combination of the two.
NOTE 3 The terms classification, regression and clustering can be characterized within the concept of tasks as
well as within the concept of methods. These terms are orthogonal both from the perspective of AI methods and
AI tasks as AI methods refer to the algorithms, such as decision trees, neural networks and k-means used to solve
problems. In contrast, AI tasks refer to the specific objectives or goals the system is trying to achieve, for example
classifying data, predicting continuous values or grouping data points. This allows for executing the same task by
different methods and using the same method for different tasks.
4.4 Understanding AI methods and AI system capabilities
An AI capability or AI component capability can be defined as the ability of an AI component to perform an
AI task. AI tasks are for instance responses and actions. Inputs and outputs of the component are both data.
Descriptors for the AI methods and AI system capabilities dimensions can be used individually as well
as in combination with each other. Distinguishing between the dimensions of AI methods and AI system
capabilities simplifies effective communication and collaboration among AI stakeholders as well as civil
society. In this context, using the dimensions enables a common language and a shared understanding

© ISO/IEC 2026 – All rights reserved
ISO/IEC DIS 42102:2026(en)
of AI systems for knowledge sharing, assessment and the development of best practices. Especially the
simultaneous consideration of both dimensions helps identify synergies and potential gaps in AI research
and development. The understanding about the choice of specific methods to realize certain capabilities
lays the basis to assess the operational effort, use case-specific dependency of AI-relevant properties as well
as the experience of stakeholders as well as representatives from civil society involved to achieve specific
goals.
4.5 Application to AI systems
Applying the descriptors in this document to an AI system involves identifying and listing the methods used
in the AI system and its resulting capabilities. Relevant stakeholders can then use the list of methods and
capabilities to:
— communicate about the AI system with a common understanding;
— determine the suitability of the AI system for a given task;
— determine risks and mitigations;
— determine an appropriate test regime and metrics;
— determine whether the system is subject to any compliance requirements;
— determine an appropriate monitoring, update and incident reporting plan;
— determine opportunities for continuous improvement;
— describe the scope of AI systems in AI management systems and AI impact assessments.
5 AI system methods dimension
5.1 General
Considering the rapid pace of development in AI technology, this document describes a variety of methods
but is not a comprehensive list of all possible AI methods. Table 1 organizes AI methods under the four
non-exclusive descriptors traditional AI, symbolic AI, machine learning and hybrid AI, each encompassing
similarly non-exclusive more specific descriptors, which can serve as orientation for the application or
[5]
description of AI system methods. Machine learning methods typically revolve around a model with
parameters that are set via a training algorithm. Inference algorithms use the fully or partially trained
model to generate outputs. Inference algorithms are sometimes used as part of training.
Table 1 provides an overview of the AI methods discussed in [6], [2], [1], [7].

© ISO/IEC 2026 – All rights reserved
ISO/IEC DIS 42102:2026(en)
Table 1 — Descriptors for AI methods
domain 1st level descriptor 2nd level descriptor
search
optimization
traditional AI
planning
decision-making
knowledge representation
logical reasoning
probabilistic reasoning
symbolic AI
non-probabilistic reasoning
further approaches
AI methods
for reasoning with uncertainty
reinforcement learning
supervised learning
machine learning unsupervised learning
semi-supervised learning
self-supervised learning
hybrid neuronal systems
hybrid AI learning with knowledge
conversational learning
Each AI method is described using descriptors on two levels. The levels can serve as orientation in the
application of this document. Due to the multi-perspective as well as subjective definitions of AI methods
in general and their corresponding changes through research results, it can be assumed that the AI method
spectrum is constantly expanding from the perspective of stakeholders.
Accordingly, methods that are not designated in the framework but are methodologically equivalent can be
characterized with appropriate descriptors.
As scientific development in the field of artificial intelligence is progressing rapidly, further methods not
described in this standard can become popular in industry and science in the future. According to the
current state of science and technology, also such currently ‘unknown methods’ can be realized using the
capabilities in Table 8. Furthermore, Annex D contains an overview of AI paradigms, features and methods.
5.2 Traditional AI
5.2.1 General
As depicted in Table 2, traditional AI can encompass algorithmic strategies for search, optimization, planning
[2]
and decision making .
NOTE In the context of this clause, the terms search, planning, decision-making and optimization refer to classes
of algorithms, i.e. they describe methods. These terms are used in this way in business communications, algorithmic
libraries and software products. Due to historical naming conventions, the same terms are sometimes also used to
describe tasks or types of tasks solved by such algorithms, i.e. these terms are sometimes used to describe capabilities.
Table 2 — Overview of traditional AI methods
traditional AI
planning and
search optimization decision-making
plan recognition
© ISO/IEC 2026 – All rights reserved
ISO/IEC DIS 42102:2026(en)
5.2.2 Search
Search algorithms systematically visit a subset of all possible states (or structures) until the goal state
(or structure) is reached. A search algorithm typically yields one or multiple goal-satisfying states (or
structures), a path through the search space to such a state (or structure) or both. On the one hand, search
algorithms encompass direct approaches with direct strategies focusing on the resolution of mathematical
formulas or their mapping to existing known problems. The resolution of mathematical formulas as such
can encompass polynomials for curve fitting for pattern identification and Sieve of Eratosthenes (e.g. for
finding prime numbers). On the other hand, search also includes search strategies, e.g. the use of heuristic
[8] [9]
search methods. , In contrast to solving Boolean satisfiability problems (SAT) and satisfiability modulo
theories (SMT) search methods are not logic-based due to the focus on exploring paths in a state space, often
using heuristics, rather than following predefined logical statements.
5.2.3 Optimization
Optimization aims to achieve the optimal solution through the use of various methods, including heuristic,
meta-heuristic, hyper-heuristic and hybrid optimization techniques. Heuristic optimization approaches
rely on domain-specific knowledge and use rule-based or iterative algorithms to guide the search process
towards the optimal solution. In contrast, meta-heuristic methods draw inspiration from natural phenomena
or social behaviour to explore the solution space. By adapting the algorithmic strategy based on problem
characteristics or performance feedback, hyper-heuristics enable algorithms to learn and evolve over time,
enhancing their ability to find optimal solutions. Additionally, hybrid optimization techniques combine
[2] [10]
different methods to leverage strengths and overcome limitations , .
5.2.4 Planning and plan recognition
5.2.4.1 General
This clause describes autonomous and semi-autonomous planning procedures as well as plan recognition.
5.2.4.2 Autonomous and semi-autonomous planning
Planning encompasses various autonomous and semi-autonomous procedures, including state-space search,
[5]
planning graphs, hierarchical planning and non-deterministic planning .
While state space search focuses on algorithm-based exploration of the space of possible states to find an
optimal solution, planning graphs capture the dependencies and relationships between actions and states
[11] [12]
based on a data structure , .
In hierarchical planning, complex tasks are broken down into smaller subtasks and organized hierarchically,
which facilitates the management of complicated tasks by breaking them down into manageable components.
With regard to unpredictable events or actions, non-deterministic planning takes uncertainties and non-
[2]
deterministic factors into account .
5.2.4.3 Plan recognition
For plan recognition, different approaches including abductive, deductive and library- and synthesis-based
methods can be used.
Abductive plan recognition focuses on proposing plausible explanations
...