oSIST prEN ISO/IEC 25029:2026

• Abstract:

  This standard applies to nudging mechanisms enhanced by AI systems.; This document provides definitions, concepts, and guidelines to address AI-enhanced nudging mechanisms by organisations. ; This standard aims to support organisations to deal with AI-enhanced nudging mechanisms in alignment with existing AI standards. ; “AI-enhanced nudging mechanisms” are a sub category of digital nudges and which are enhanced by AI systems. ; It provides use-cases to illustrate AI-enhanced nudging mechanisms. It provides guidelines and requirements for designing responsible AI-enhanced nudging mechanisms. This includes horizontal processes and key indicators using specific vertical examples.

• Status: Not Published
• Public Enquiry End Date: 07-Sep-2026
• 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 25029 - Artificial intelligence - AI-enhanced nudging (ISO/IEC DIS 25029:2026)

Overview

oSIST prEN ISO/IEC 25029:2026: Artificial intelligence - AI-enhanced nudging (ISO/IEC DIS 25029:2026) sets out definitions, concepts, use cases, and guidelines for organizations implementing or using AI-enhanced nudging mechanisms. Developed in collaboration with CEN, ISO, and IEC, this draft international standard addresses the unique aspects and ethical considerations of using artificial intelligence to shape digital nudging techniques within various applications. It supports organizations in aligning AI-enhanced nudging practices with established AI standards to ensure responsible, transparent, and ethical use, particularly important as digital influence increases across industries and societal domains.

Key Topics

AI-enhanced nudging refers to digital nudges-subtle changes in a digital choice architecture-augmented through AI systems. The standard covers:

• Definitions and concepts: Clarifies terms such as nudging, digital nudge, sludge, choice architecture, bias, human cognitive bias, affective computing, and persuasive technology.
• Types of AI-enhanced nudging mechanisms:
  • AI-triggered nudges: AI selects and activates nudges based on user data profiles.
  • AI-sorted nudges: AI determines sequencing and timing of nudges for maximum effect.
  • AI-generated nudges: AI dynamically creates or adjusts nudges (e.g. interface colors, tone, content), often personalizing in real time.
• Ethical design and risk: Identifies processes for ethical evaluation (choice architecture analysis, ethics risk assessment, harm mitigation, and transparency).
• Horizontal processes and key indicators: Offers methodologies that are technology-agnostic, applicable across sectors and adaptable to future innovations.
• Impact on vulnerable groups: Provides insights on age-appropriate policies and the importance of protecting consumers and children.

Applications

The guidance in oSIST prEN ISO/IEC 25029:2026 is valuable to a broad spectrum of organizations and sectors employing or regulating AI-enhanced digital nudges, including:

• Health: Personalized reminders, adaptive health tracking notifications, and subtle prompts for healthy behaviors.
• Education: Custom learning prompts, reinforcement strategies, and engagement nudges tailored through AI insights.
• E-commerce and marketing: Personalized product recommendations, targeted offers, and interface adjustments that influence purchasing decisions.
• Social networks and communication platforms: Content sequencing, opinion bubble filtering, and notification management driven by AI to modulate engagement.
• Cybersecurity: Real-time, individualized warning systems or prompts, encouraging secure behaviors among end-users.
• Smart devices and IoT: Adaptive interfaces and behavioral prompts in smart home systems, vehicles, and wearables.

By following these guidelines, organizations can create responsible AI-driven nudging mechanisms that promote user well-being, transparency, and trust, while minimizing ethical risks and the potential for negative, unintended consequences such as manipulation or discrimination.

Related Standards

This document aligns with and complements other important AI and digital ethics standards, including:

• ISO/IEC 22989: Artificial intelligence concepts and terminology.
• ISO/IEC 24027: Bias in AI systems and AI-augmented decision-making.
• ISO/IEC 42001: Artificial intelligence - Management system.
• ISO/IEC TS 6254: Explainability of AI systems.
• ISO/IEC TR 24028: AI trustworthiness considerations.
• ISO/IEC 30150-1: Affective computing.
• ISO/IEC 38500: Governance of IT and AI.
• ISO 31000: Risk management.

Organizations and professionals developing, deploying, or overseeing AI-powered digital experiences are encouraged to use oSIST prEN ISO/IEC 25029:2026 in conjunction with these standards to foster a robust framework for ethics, transparency, and user empowerment in AI-enhanced environments.

---

By embedding the principles and methodologies outlined in oSIST prEN ISO/IEC 25029:2026, organizations can proactively address the opportunities and risks of AI-enhanced nudging mechanisms, ensuring their AI strategies are both effective and ethically grounded.

SLOVENSKI STANDARD
01-september-2026
Umetna inteligenca - Izboljšano usmerjanje z UI (ISO/IEC DIS 25029:2026)
Artificial intelligence - AI-enhanced nudging (ISO/IEC DIS 25029:2026)
Künstliche Intelligenz - KI-gestütztes Nudging (ISO/IEC DIS 25029:2026)
Intelligence artificielle - Nudges renforcés par l’IA (ISO/IEC DIS 25029:2026)
Ta slovenski standard je istoveten z: prEN ISO/IEC 25029
ICS:
35.240.01 Uporabniške rešitve Application of information
informacijske tehnike in technology in general
tehnologije na splošno
2003-01.Slovenski inštitut za standardizacijo. Razmnoževanje celote ali delov tega standarda ni dovoljeno.

DRAFT
International
Standard
ISO/IEC DIS 25029
ISO/IEC JTC 1/SC 42
Artificial intelligence — AI-
Secretariat: ANSI
enhanced nudging
Voting begins on:
Intelligence artificielle — Nudges renforcés par l’IA
2026-06-15
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 25029:2026(en)
DRAFT
ISO/IEC DIS 25029:2026(en)
International
Standard
ISO/IEC DIS 25029
ISO/IEC JTC 1/SC 42
Artificial intelligence — AI-
Secretariat: ANSI
enhanced nudging
Voting begins on:
Intelligence artificielle — Nudges renforcés par l’IA
ICS: 35.240.01 Voting terminates on:
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 25029:2026(en)
© ISO/IEC 2026 – All rights reserved
ii
ISO/IEC DIS 25029:2026(en)
Contents Page
Foreword .iv
1 Scope .1
2 Normative references .1
3 Terms and definitions .1
3.1 General and decision concepts .1
3.2 Ethics concepts .3
3.3 Terms related to risk and quality .4
3.4 Vulnerability terms .6
4 Acronyms . 6
4.1 AI .6
4.2 CRIA . .6
4.3 XAI .7
5 AI-enhanced nudging mechanisms .7
5.1 Introduction .7
5.2 AI-enhanced nudging concept .7
5.3 Types of AI-enhanced nudging mechanisms .8
5.4 Methodology .9
5.5 AI-enhanced nudging as informational perturbation mechanisms .9
5.6 Strategy for processing AI-enhanced nudging mechanisms .11
6 Process and criteria for handling AI-enhanced nudging mechanisms .12
6.1 Introduction . 12
6.2 Description of the AI system, scope and targets . 12
6.3 Ethical design . 12
6.4 Nudge detection process. 13
6.5 Choice architecture .14
6.6 Multidisciplinary risk identification .17
6.7 Informational dimensions analysis .18
6.8 Ethical foresight analysis .19
6.9 Declaration and strategy for AI-enhanced nudges .19
6.10 Ethics risk assessment . 20
6.11 Threshold of ethical neutrality . 20
6.12 Components of an ethics risk assessment .21
6.13 Declaration of ethical neutrality . 22
6.14 Monitoring mechanisms . 22
6.15 Mitigation strategies . 23
Annex A (informative) Useful guidances .26
Annex B (informative) AI-enhanced nudges use cases .29
Bibliography .42

© ISO/IEC 2026 – All rights reserved
iii
ISO/IEC DIS 25029:2026(en)
Foreword
ISO (the International Organization for Standardization) is a worldwide federation of national standards
bodies (ISO member bodies). The work of preparing International Standards is normally carried out through
ISO technical committees. Each member body interested in a subject for which a technical committee
has been established has the right to be represented on that committee. International organizations,
governmental and non-governmental, in liaison with ISO, also take part in the work. ISO collaborates closely
with the International Electrotechnical Commission (IEC) on all matters of electrotechnical standardization.
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 ISO documents 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).
Attention is drawn to the possibility that some of the elements of this document may be the subject of patent
rights. ISO 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).
Any trade name used in this document is information given for the convenience of users and does not
constitute an endorsement.
For an explanation on 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 the following URL: www.iso.org/iso/foreword.html.
This document was prepared by 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.

© ISO/IEC 2026 – All rights reserved
iv
DRAFT International Standard ISO/IEC DIS 25029:2026(en)
Artificial intelligence — AI-enhanced nudging
1 Scope
This document provides definitions, concepts, guidelines and methodology to address AI-enhanced nudging
mechanisms by organizations. It provides requirements for designing responsible AI-enhanced nudging
mechanisms, key indicators, both horizontally and vertically. This document supports organizations that
develop or use AI-enhanced nudges, as well as entities interested in the protection of civil society and
individuals, including consumers and workers.
2 Normative references
There are no normative references in this document.
3 Terms and definitions
For the purposes of this document, the following terms and definitions apply.
ISO and IEC maintain terminological databases for use in standardization at the following addresses:
— IEC Electropedia: available at http:// www .electropedia .org/
— ISO Online browsing platform: available at http:// www .iso .org/ obp
3.1 General and decision concepts
3.1.1
choice architecture
organizing the context in which people make decisions
Note 1 to entry: The choice architecture is designed by choice architects (human agents) or inferred by AI systems
(artificial agents).
3.1.2
nudge
nudging mechanism
subtle change in the choice architecture to encourage shifts in behaviour based on human cognitive bias
(3.1.11)
[1]
[SOURCE: ISO/IEC TR 24027:2021 ]
3.1.3
sludge
specific instance of nudge that adds friction in making a choice based on biases
3.1.4
digital nudge
nudge that is delivered in digital environment or using digital technologies

© ISO/IEC 2026 – All rights reserved
ISO/IEC DIS 25029:2026(en)
3.1.5
affective computing
collection, recognition, strategy and presentation of affective characteristics of human-computer
interactions
[2]
[SOURCE: ISO/IEC 30150-1:2022 , 3.2]
3.1.6
persuasive technology
technology that is designed to change opinions, attitudes or behaviours of the users through persuasion and
social influence
3.1.7
artificial intelligence system
engineered system that generates outputs such as content, forecasts, recommendations or decisions for a
given set of human-defined objectives
Note 1 to entry: The engineered system can use various techniques and approaches related to artificial intelligence to
develop a model to represent data, knowledge and processes which can be used to conduct tasks.
Note 2 to entry: AI systems are designed to operate with varying levels of automation.
[3]
[SOURCE: ISO/IEC 22989:2022 ]
3.1.8
AI-enhanced nudges
AI nudges
AI-enhanced nudging mechanisms
use of artificial intelligence techniques to design and deliver behavioural interventions that subtly influence
individuals' decisions or actions
3.1.9
implicit AI-enhanced nudging
AI-enhanced nudging which is not intended by the organisation that provides the nudging mechanism
3.1.10
bias
systematic difference in treatment of certain objects, people or groups in comparison to others
Note 1 to entry: Treatment is any kind of action, including perception, observation, representation, prediction or
decision.
[1]
[SOURCE: ISO/IEC TR 24027:2021 , 3.2.2]
3.1.11
human cognitive bias
bias that occurs when humans are processing and interpreting information
Note 1 to entry: Human cognitive bias influences judgement and decision-making.
[1]
[SOURCE: ISO/IEC TR 24027:2021 , 3.2.4]
3.1.12
human factors
environmental, organizational and job factors, in conjunction with cognitive human characteristics, which
influence the behaviour of persons or organizations
[4]
[SOURCE: ISO/IEC TR 24028:2020 , 3.19]
3.1.13
neurodata
neural network patterns and other biological signals involved in cognitive and neural functioning

© ISO/IEC 2026 – All rights reserved
ISO/IEC DIS 25029:2026(en)
3.2 Ethics concepts
3.2.1
ethics
branch of philosophy that analyses and reflects on moral situations, including the principles, values and
frameworks involved
3.2.2
code of ethics
set of principles and rules concerning moral obligations and regard for the rights of humans and nature
Note 1 to entry: Codes of ethics can be specified by a given profession, group or organization.
3.2.3
code of data ethics
set of ethics guidelines, principles and procedures by which data is acquired, analysed, processed, adjusted,
compiled or otherwise sold, traded or shared with other entities
3.2.4
ethics committee
group of individuals with multidisciplinary expertise responsible for reviewing, advising on and overseeing
the ethical aspects of the organization's activities, projects or policies and requiring the participation of at
least one professional ethicist
3.2.5
professional ethicist
a professional with competence in studying, analysing and providing guidance on the ethical considerations,
implications and practices
3.2.6
ethics risk assessment
process of identifying, analysing and evaluating potential ethical risks associated with a system, action or
decision
3.2.7
ethical risk
potential for an action, system or decision to affect ethical objectives
3.2.8
ethical harm
adverse effect on individuals, communities or societal values that occur via breaches of ethical principles
stated in a code of ethics (3.2.2)
3.2.9
ethical foresight analysis
prediction of potential ethical issues in new technologies that guide proactive measures to address these
concerns
3.2.10
explainability
property of an AI system (3.1.7) that enables a given human audience to comprehend the reasons for the
system's behaviour
Note 1 to entry: Explainability methods are not limited to the production of explanations, but also include the enabling
of interpretations.
[5]
[SOURCE: ISO/IEC TS 6254:2025 ]
3.2.11
explainable AI
approach to explain information about the functioning or output of an AI system

© ISO/IEC 2026 – All rights reserved
ISO/IEC DIS 25029:2026(en)
3.2.12
moral situation
circumstance in which an agent faces a decision requiring evaluation of actions based on moral principles,
values or obligations
3.2.13
semantic capital
any content that can enhance someone’s power to give meaning to and make sense of (semanticise) something
Note 1 to entry: Large-scale informational perturbations can erode the shared content that enables individuals and
societies to make coherent sense of reality, thereby undermining meaning, trust and collective understanding; for
example, algorithmically personalised nudges that systematically amplify misleading or inconsistent narratives can
degrade the semantic resources upon which public debate and democratic decision-making rely.
3.2.14
shared moral framework
common ethics-based principles and values guiding an organization's decisions and behaviours
3.2.15
well-being
fulfilment of physical, mental and cognitive needs and expectations
Note 1 to entry: Well-being relates to all aspects of life.
Note 2 to entry: Well-being exists at the individual, household, country and global level and can be applied to people
and nature and to individuals and systems.
[6]
[SOURCE: ISO/UNDP PAS 53002:2024 , 3.38]
3.2.16
infosphere
environment consisting of all informational entities, processes and interactions, including both digital and
analogue components, in which information is created, stored, exchanged or used, where humans, artificial
agents and systems coexist and interact
3.2.17
ethics standard
value or principle that guides judgements of right and wrong and informs responsible conduct within a given
context
3.3 Terms related to risk and quality
3.3.1
management system
set of interrelated or interacting elements of an organization (3.3.8) to establish policies and objectives,
aswell as process (3.3.2) to achieve those objectives
Note 1 to entry: A management system can address a single discipline or several disciplines.
Note 2 to entry: The management system elements include the organization’s structure, roles andresponsibilities,
planning and operation.
[7]
[SOURCE: ISO/IEC 42001:2023 , 3.4]
3.3.2
process
set of interrelated or interacting activities that uses or transforms inputs to deliver a result
[7]
[SOURCE: ISO/IEC 42001:2023 , 3.8]

© ISO/IEC 2026 – All rights reserved
ISO/IEC DIS 25029:2026(en)
3.3.3
AI system impact assessment
impact assessment
formal, documented process by which the impacts to individuals, groups of individuals and societies are
considered by an organization developing, providing or using products or services utilizing artificial
intelligence
[8]
[SOURCE: ISO/IEC 42005:2025 , 3.1]
3.3.4
child rights impact assessment
process to identify, analyse and evaluate the actual or potential impacts of a policy, project, product, service,
system or activity on the rights and well-being of children, in order to inform decision-making and support
the prevention of adverse impacts
Note 1 to entry: The CRIA is one of the general measures of implementation of the United Nations Convention on the
Rights of the Child (UNCRC), according to United Nations Children’s Fund (UNICEF).
3.3.5
governance
human-based system comprising directing, overseeing and accountability
[9]
[SOURCE: ISO/IEC 38500:2024 , 3.3]
3.3.6
goal
intended outcome
[10]
[SOURCE: ISO/IEC 25022:2016 ]
3.3.7
hazard
potential source of harm
[11]
[SOURCE: ISO/IEC Guide 51:2014 , 3.5]
3.3.8
organization
person or group of people that has its own functions with responsibilities, authorities and relationships to
achieve its objectives
Note 1 to entry: The concept of organization includes, but is not limited to, sole-trader, company, corporation, firm,
enterprise, authority, partnership, charity or institution or part or combination thereof, whether incorporated or not,
public or private.
[7]
[SOURCE: ISO/IEC 42001:2023 , 3.1]
3.3.9
risk
effect of uncertainty on objectives
[12]
[SOURCE: ISO/IEC 23894:2023 XX]
3.3.10
risk identification
the process of finding, recognising and recording risks
[13]
[SOURCE: ISO 31000:2018 , 2.15]

© ISO/IEC 2026 – All rights reserved
ISO/IEC DIS 25029:2026(en)
3.3.11
residual risk
risk remaining after risk treatment (3.3.12)
[14]
[SOURCE: ISO Guide 73:2009 XX]
3.3.12
risk treatment
process to eliminate risk or reduce it to a tolerable level
[14]
[SOURCE: ISO Guide 73:2009 ]
3.3.13
task
set of activities undertaken to achieve a specific goal
[15]
[SOURCE: EN ISO 9241-11:2018 , 3.1.11]
3.4 Vulnerability terms
3.4.1
age group
children who fall within a defined age range linked to specific developmental milestones or characteristic
behaviours
[16]
[SOURCE: ISO/TR 8124-8:2024 , 3.1]
3.4.2
child
children
person below the age of 18 years
Note 1 to entry: As defined in the United Nations Convention on the Rights of the Child (UNCRC) and referred to in
International Labour Organization (ILO) Convention 182[9].
Note 2 to entry: National applicable statutory or regulatory requirements may define a different age limit for a child.
[17]
[SOURCE: IWA 49:2025 ]
3.4.3
age-appropriate policy
an organization's position outlining its commitment to age-appropriate experience, disclosure and consent,
including the identification of target age ranges
3.4.4
protected category
category defined under law or regulation by jurisdiction, that can include race, age, gender, religion, ability/
disability, sexual orientation, creed, skin colour, nation of origin, socio-economic factors
4 Acronyms
For the purposes of this document, the following acronyms apply.
4.1 AI
artificial intelligence
4.2 CRIA
child rights impact assessment

© ISO/IEC 2026 – All rights reserved
ISO/IEC DIS 25029:2026(en)
4.3 XAI
explainable AI
5 AI-enhanced nudging mechanisms
5.1 Introduction
The concept of nudging was developed and disseminated in behavioural economics to improve the well-
being of society through individual decision-making incentives. A nudge is a subtle incentive in the design of
a digital system to drive behaviour. Nudges should be distinguished from business analytics: while nudges
subtly influence individual behaviour using behavioural insights without limiting choice, business analytics
relies on data-driven decisions to optimise performance. Nudges can also be designed to improve an
organization performance, making them an instrument of analytics-driven strategy. Nudges can be physical,
such as for the placement of healthy food in a canteen or digital ones ranging from user interface nudges
such as the position of a button on a web page to sophisticated nudges enhanced by AI such as a personalised
news or product recommendations feed. At present, digital nudge mechanisms can be merged into online
technologies, such as email, pop-ups, short message service (SMS), web interfaces, smart watches, mobile
apps, home appliances, smart cars, chatbots, robots and video games. Although they can take many forms,
these digital nudges are often implemented as reminders and alerts, pre-selected default settings (e.g.
subscriptions, renewals) or interface design features that make a preferred option more convenient, salient
or accessible. They can be intentionally and explicitly built into design as part of choice architecture, but can
also arise unintentionally and implicitly within interface design such as with the positioning or sequencing
of options. However, the intrinsic nature of a nudge changes when a digital nudging mechanism is enhanced
by AI systems (so-called AI-enhanced nudging mechanism).
AI-enhanced nudges operate by leveraging vast amounts of data and exploiting human factors such as
cognitive biases, emotional impulses and other behavioural mechanisms, forming a type of persuasive
[18]
technology studied within the field of affective computing . The use of AI-enhanced nudges is a powerful
instrument for organizations; however, the effect of these nudges on individuals and societies raises ethical
concerns. A nudge can be morally loaded based on its consequences, the motivation that brought it about
or its effect on the well-being of the affected or potentially affected person or group of persons. Research
has shown that the public tends to be more accepting of policy-related nudges that can support choices
aligned with individual interests and values, such as nudges that highlight calories (kilojoules) in fast food
restaurants, but less approving of nudges that limit or remove choice, such as with default enrolment into
programs or promotions that require opting out. Importantly, when evaluating any morally loaded action,
it is essential to distinguish between actions that have moral relevance and actions that can be considered
morally neutral or in a state of moral inertia, where no positive or negative outcome can be reasonably
anticipated.
In 5.2, the abstract idea or general notion of AI-enhanced nudging is presented. In 5.3, the specific process
or system by which the concept of AI-enhanced nudges is implemented or functions is detailed. In 5.3, the
different types of AI-enhanced nudges are described. In 5.4, the methodology for identifying and managing
AI-enhanced nudges is outlined. In 5.5, AI-enhanced nudging mechanisms are analysed as informational
disruptions that can unpredictably influence decisions, requiring ethical evaluation and risk mitigation. In
5.6, the strategy for processing AI-enhanced mechanisms is defined.
5.2 AI-enhanced nudging concept
Initially, a nudge was defined as “any aspect of the choice architecture that alters people’s behaviour in a
[19]
predictable way without forbidding any options or significantly changing their economic incentives” . The
evolution of nudges on digital interfaces can be described as a specific instance of an imperceptible incentive
in the design of the choice architecture that uses digital systems or interfaces to drive behaviour. Choice
architects create those individual mechanisms to interfere with user behaviour during an interaction on any
connected device, from phones and wearables to smart homes, cars, health trackers and robots. Practically,
digital nudges can be embedded in visual interfaces (e.g. buttons, colours, robot facial expressions), sound
interfaces (e.g. voices, music, noises), tactile interfaces (e.g. vibrations) or brain computer interfaces. They
can shape available information such as sequences (e.g. display of products order), filtering (e.g. social

© ISO/IEC 2026 – All rights reserved
ISO/IEC DIS 25029:2026(en)
network opinion bubbles) for triggering emotions or to influence the user’s behaviour. Nowadays, digital
nudging mechanisms are used positively in several domains: health, education, gaming, social networks and
cybersecurity.
AI systems, with their ability to detect user’s behaviours and attention based on big data analytics, are
designing personalised interfaces and experiences in a relatively autonomous way to influence a user's
intentions through nudge strategies, thereby transforming individual nudging mechanisms into complex
socio-technical systems. Thus, while digital nudges can be subtle rule-based design elements, AI-enhanced
nudges are dynamic and adaptive and can be personalised in real time via user data.
Finally, AI-enhanced nudges are not to be confused with subliminal techniques. AI-enhanced nudging can
influence behaviour both consciously and non-consciously, depending on the design and context of use. Some
techniques can approach or even cross the threshold of subliminal influence, thereby raising concerns about
manipulation and autonomy. Distinguishing between acceptable and unacceptable forms of behavioural
influence should rely on empirical evidence and ethical criteria, not on an assumed awareness threshold.
However, improperly applied digital nudging techniques can also cause harm where users are nudged
towards decisions that are not in their best interests, particularly among vulnerable groups such as children
who are generally less able to resist social pressures.
In general, AI-enhanced nudging mechanisms are personalisations of the nudging strategies that can perturb
the decision-making process.
AI-enhanced nudging mechanisms can be created by modifying a digital nudge or a sequence of known
digital nudges (taken from a repository of nudges) or AI systems can attempt to optimise performance by
varying and combining interface or content modifications in unpredictable ways. In the first case, some
consequences of AI-enhanced nudging mechanisms are not caused by a single digital nudge but are generated
by the combination (sequence and classification), frequency (cadence and intensity) and distribution
(volume of receivers) of nudges. In the second case, the consequences of nudging mechanisms can be
unpredictable because they are generated by the AI system’s ability to adapt behaviour to user interactions.
In this situation, the effects of the mechanism can only be monitored and mitigated after the consequences
appear, sometimes without being able to isolate each individual mechanism that makes up the flow aimed at
[20][21]
inciting a change in the decision-making process.
5.3 Types of AI-enhanced nudging mechanisms
[22]
Basic elements include digital nudges , which are subtle prompts or reminders designed to influence user
behaviour in a particular direction. They are typically delivered through digital devices or platforms, such as
smartphones or websites and can take a variety of forms, from choice elements (e.g. buttons) to notifications
(e.g. pop-ups). This describes the basic elements of digital nudges and the three ways of instantiating AI-
enhanced nudging mechanisms.
AI-enhanced nudging mechanisms are composed of digital nudges that can be activated, reorganized or
entirely generated by AI systems. Yet, AI-enhanced nudging mechanisms can be the unintended result of a
process of actions that end up impacting the decision-making process.
Therefore, three types of AI-enhanced nudging mechanisms can be identified, consisting of sets of digital
nudges.
Instantiation of AI-enhanced nudging mechanisms can be done in the following:
— AI system triggers a nudge (or a set of nudges): The AI system decides for itself when to use nudges or
a preset set of nudges, deciding which user profile and with what frequency to distribute the nudges to
users in order to benefit the AI provider.
— AI system sorts a set of nudges: The AI system determines in which order to distribute the available
nudges. For example, a sequence of nudges can be organized and adapted to the user in order to maximise
the AI provider's profits.
— AI system generates nudges on interfaces: The peculiarity of AI-generated nudges lies in the pervasiveness
with which they affect the user’s decision-making autonomy. For example, micro variables (e.g. interface

© ISO/IEC 2026 – All rights reserved
ISO/IEC DIS 25029:2026(en)
colour, voice tone and syntactic sentence structure) can be modified slightly to influence the user’s
behaviour (i.e. staying longer on an interface) with the aim of benefiting the AI provider’s objectives (i.e.
making the user spend more).
AI-enhanced nudging can involve one or a combination of more distinct mechanisms.
5.4 Methodology
This clause identifies observables to establish process and criteria to mitigate ethics risks and consequently
improve the trustworthiness of the nudging solution.
Nudges are generally seen as mechanisms that push individuals toward specific outcomes (e.g. adopting
[19]
healthier behaviours) , while, on the contrary, sludges add friction to achieving specific outcomes (e.g.
[23]
preventing service unsubscription) . This document adopts a neutral approach, using the term “nudges”
to refer to both slight pushes and frictions.
Using the informational approach, which treats all elements of the interaction process as informational
entities, AI-enhanced nudging mechanisms can be defined as perturbation mechanisms of the information
process that have no a priori moral charge unless the trade-off of the AI provider’s advantages is explicitly
against the users. The intentionality of the goal-setting phase of the system is thus just as important as
monitoring hazards and mitigating ethical harms.
Considering nudging mechanisms as information elements has a twofold relevance. Firstly, the nudging
mechanisms can be treated as technology-agnostic mechanisms, so this document can apply even if future
nudging innovations emerge. Secondly, the information elements can be used to highlight the dimensions of
informational interaction between the various elements, thereby bringing to light informational observables,
such as the semantic level of receivers, agents or mediators. For example, the age-appropriate policy for
minors can be a simple way to evaluate the semantic (cognitive, psychological) capabilities of receivers.
Understanding the semantic level of the involved elements allows the construction of better-performing
monitoring mechanisms 6.14 and mitigation mechanisms 6.15.
A nudge should be analysed with an appropriate level of abstraction (conceptual viewpoint) and granularity
(detail). The appropriate level depends on the intended purpose of the analysis (e.g. ethical evaluation,
technical implementation, legal compliance). To ensure meaningful and transparent insights, analysts must
explicitly define and justify their chosen levels, guided by clear criteria such as audience, regulatory context
or the system’s ethical harms.
For each phase of the process outlined in 5.6, the stakeholders can use the questioning structure presented
below. The questions should be seen as a baseline upon which more advanced analysis can be conducted.
a) The why questions seek to clarify the importance and purpose of the phase.
b) The what questions determine the specific elements or factors that need attention during this phase.
c) The who questions identify the key individuals or teams responsible for or need to be part of this phase.
d) The where questions specify the location, either physical or digital, where this phase should take place
or be focused on.
e) The when questions set the timing for initiating the process.
f) The how questions describe the methodology or approach to be followed in this phase.
g) The output enumerates the outcomes or results expected from completing this process phase.
This structure ensures a comprehensive understanding and efficient execution of each phase, from
conceptualisation to realisation, ensuring all critical aspects are addressed.
5.5 AI-enhanced nudging as informational perturbation mechanisms
AI-enhanced nudging mechanisms intervene to influence the decision-making process of at least one user,
towards one decision instead of another. The choice of this disruption becomes ethically critical when the

© ISO/IEC 2026 – All rights reserved
ISO/IEC DIS 25029:2026(en)
advantages of the choice are to the detriment of the user or the society to the point of triggering ethical
harms or eliciting threats.
AI-enhanced nudges are regarded as informational perturbation mechanisms that interfere with the
moral situation where an agent triggers the informational process toward a receiver. A moral situation is a
condition in which a moral action takes place. All interactions between an agent and a receiver generate a
moral situation, yet most situations exist in a state of moral inertia. Their consequences are neutral. At the
same time, others can cause positive consequences or negative consequences. From an ethical perspective,
the aim of this document is to define the mechanisms and dimensions to be taken into account in order to
understand how to monitor or mitigate the ethical harms of an AI-enhanced nudging mechanism.
The process of analysing the elements that constitute the moral situation allows for examining various
dimensions of any hazards. From this perspective, dangers are not limited to individuals but also their
relationships and social contracts. For individuals, AI-enhanced nudging mechanisms can be regarded as
risk-increasing or risk-triggering elements. For social relationships, AI-enhanced nudging mechanisms can
be regarded as elements that increase or trigger ethical harms.
The ethical harms that shall be taken into account are based on the harm hierarchy used in the medical
[24] [25][26]
device sector , such as injury or damage to people's health or damage to property or the
environment. Recognising that a list of ethical harms can be sectorially adapted, the list below proposes an
example of a hierarchy of ethical harms in order of severity of impacts that goes from low adverse impact
(petty disturbance) to critical adverse impact (life or death decisions):
a) life or death decisions;
b) physical or mental harm;
c) loss of rights or freedoms;
d) restrictions of rights or freedoms;
e) restriction to the access of goods and services;
f) discriminatory outcomes, where not covered above;
g) unfair outcomes;
h) identity theft or loss of identity;
i) disclosure of personal information;
j) damage to reputation;
k) monetary loss;
l) harassment and increased undesirable exposure;
m) annoyance and hassle;
n) minor repairs;
o) petty disturbance.
The societal and political threats that should be taken into account are based on internationally recognised
[27][28]
guidelines or scientific literature. The list below shows the threats that cannot be prioritized without
contextual analyses:
p) freedom of choice;
q) group privacy;
r) freedom of opinion;
s) discrimination;
© ISO/IEC 2026 – All rights reserved
ISO/IEC DIS 25029:2026(en)
t) collapse of essential services;
u) fake news;
v) digital continuity;
[29]
w) semantic capital .
Both dimensions, individual and societal, should be taken into account when analysing possible ethical
harms and threats.
5.6 Strategy for processing AI-enhanced nudging mechanisms
The strategy for handling AI-enhanced mechanisms is outlined in Figure 1 and briefly described in 5.6. The
detailed requirements will be presented in Clause 6.
Figure 1 — Strategy for processing AI-enhanced nudging mechanisms
At the outset of the process for managing AI-enhanced nudging mechanisms, the AI nudging process is
outlined in terms of description, scope and target of nudges (-a). This shall comply with an established
ethical design (-b) that can be supported by regional values (-c), a shared moral framework (-d), a code of
ethics (-e), a code of data ethics (-f) and a specialized ethics committee (-g). A nudge detection process (-h)
is used to identify implicit AI-enhanced nudges that may have arisen as part of the decision-making process
or through interface design modification. Implicit nudges are explicated and along with explicitly designed
AI-enhanced nudges are used to articulate the choice architecture (-i).
An ethics risk assessment (A) begins with a multidisciplinary risk identification (-j) tailored to the specific
architecture. It also includes an informational dimension analysis (-k) to find unknown ethical harms and an
ethical foresight analysis (-l) to predict future ethical adverse effects. This process helps to set the threshold
for ethical neutrality of adverse effects (-n) based on the description, the scope, the targets of the nudges
(-a) and the choice architecture (-i). Next, it identifies which ethical risks (-m) shall be addressed. Finally,
it documents an objective list of pros and cons for using the nudge (-o), notes the trade-offs and rationale
for the choice (-p), records any moral deliberation (-q), highlights any ethical residual risk (-r) and tracks
implementation of each moral deliberation (-s).

© ISO/IEC 2026 – All rights reserved
ISO/IEC DIS 25029:2026(en)
The ethics risk assessment is used to develop a declaration and strategy for AI nudges (C) which include the
following: the mechanisms for monitoring for ethical hazards and residual risks (t), the impacts on protected
categories of persons and the mec
...