EvalCommunity Case Study

UNESCO Ethical Impact Assessment: A Practical Framework for Responsible AI

A practical guide for development professionals, evaluators, policymakers, researchers, and organisations assessing the responsible use of AI systems.

Case Study Summary

UNESCO’s Ethical Impact Assessment, also called the EIA, is a comprehensive tool designed for policymakers and other stakeholders to examine whether a specific AI system is in line with the values and principles of UNESCO’s Recommendation on the Ethics of Artificial Intelligence.

The EIA helps teams assess potential positive and negative impacts, identify and engage relevant stakeholders, and develop mitigation strategies to address adverse outcomes.

Why this matters for development professionals

Development professionals increasingly work in environments where AI systems may support programme design, data analysis, reporting, service delivery, procurement, knowledge management, learning, and decision-making. These uses can create opportunities, but they also require careful attention to ethics, rights, accountability, fairness, transparency, privacy, and stakeholder trust.

The UNESCO EIA is useful because it gives teams a structured way to assess a specific AI system before or during use. It helps move the conversation from general interest in AI to practical questions about purpose, proportionality, stakeholder impact, safeguards, and mitigation.

Core question

Can this specific AI system be used responsibly, proportionately, transparently, and with appropriate safeguards?

What is the UNESCO Ethical Impact Assessment?

The UNESCO EIA is a tool for assessing whether a specific AI system aligns with UNESCO’s Recommendation on the Ethics of Artificial Intelligence. It supports ethical reflection across the AI lifecycle, including research, design, development, deployment, use, maintenance, operation, monitoring, and termination.

UNESCO describes the EIA as a tool for “ethics by design.” This means ethical questions should not be left until the end of a project. They should be considered throughout the lifecycle of an AI system.

Who can use the EIA?

Government officials

Including teams involved in the procurement or deployment of AI systems.

AI developers and private companies

Organisations seeking to develop or deploy AI systems ethically and in line with UNESCO’s Recommendation on the Ethics of AI.

Researchers and academics

Those analysing AI systems to ensure they meet ethical guidelines and contribute positively to society.

How development professionals can use the EIA

Development professionals can use the EIA as a practical governance and reflection tool whenever AI systems are being considered, procured, piloted, deployed, or reviewed in development and humanitarian contexts.

1. Programme design

Use the EIA to check whether an AI system has a clear purpose, whether non-AI alternatives were considered, and whether the system is proportionate to the programme objective.

2. Procurement

Use the EIA to support supplier selection and define ethical standards, safeguards, documentation requirements, and accountability expectations before procurement decisions are finalised.

3. Monitoring and evaluation

Use the EIA to assess how AI may affect evidence quality, stakeholder voice, privacy, transparency, fairness, human review, and accountability in M&E workflows.

4. Stakeholder engagement

Use the EIA to identify who may be affected by the AI system and how their perspectives will be included before and during implementation.

5. Data governance

Use the EIA to review personal data, data minimisation, consent, privacy by design, third-party access, dataset transparency, and ethical sourcing.

6. Learning and adaptation

Use the end-of-assessment reflection to identify practical next steps and support continuous improvement after the initial assessment.

When should the EIA be conducted?

UNESCO explains that the EIA is especially valuable early in the AI lifecycle. This includes project research, development and design, procurement, and initial deployment.

Project research, development and design: to reflect on scope, legitimacy, and alignment with ethical principles.
Procurement: to assist in selecting suppliers and establishing ethical standards through contractual obligations.
Initial deployment: to evaluate the AI system as it is implemented and identify areas for improvement.

Why conduct the EIA?

UNESCO states that conducting the EIA demonstrates a commitment to ethical AI. The process helps teams understand the ethical implications of an AI system, make informed choices, improve system design, build trust among users and stakeholders, and prepare for existing or forthcoming requirements for ethical impact assessments.

For development organisations, this is important because AI systems may affect communities, public services, programme decisions, evidence processes, institutional accountability, and stakeholder trust.

Structure of the UNESCO EIA

UNESCO states that the EIA has three main parts.

1. Scoping questions

These assess the fundamentals of the AI project, including project scope, objectives, and stakeholder engagement.

2. Implementing UNESCO principles

This part evaluates whether design, development, and deployment align with UNESCO’s ethical principles, with attention to procedural safeguards and mitigation strategies.

3. Mapping potential impacts

This part identifies and assesses potential positive and negative impacts across principles and criteria.

How the EIA can be completed

UNESCO states that the Ethical Impact Assessment is free to download and use. It can be completed asynchronously by team members or through one or two half-day collaborative workshops.

UNESCO recommends a collaborative approach involving diverse roles and team members to capture comprehensive insights and support actionable outcomes.

Six key stages of the EIA process

1. Identify a project manager

Assign a clear point of contact to coordinate the EIA process, including planning, timelines, and team coordination.

2. Assemble a diverse team

UNESCO notes that the team may include organisational authorities or leaders, technical developers and teams, procurement officers, legal experts, social scientists and ethicists, stakeholder engagement specialists, and change management professionals.

3. Work independently on the EIA

Team members may respond to or review sections asynchronously, based on their expertise or role in the AI lifecycle.

4. Facilitate a collaborative workshop

A workshop can be used to review difficult procedural questions and complete impact mapping tables through collaborative discussion.

5. Engage stakeholders

UNESCO encourages a broad range of stakeholders to be engaged throughout the EIA process, especially from the earliest stages.

6. Complete the end-of-assessment reflection

The final stage invites the team to reflect on the process, identify practical next steps, and support continuous improvement.

What development teams should prepare before using the EIA

Before completing the EIA, development teams can prepare basic information about the AI system and its intended context of use. This helps the assessment become more practical and evidence-based.

Name or working title of the AI system.
Purpose of the AI system.
Project owner or responsible team.
Lifecycle stage of the system.
Whether the system is being procured.
Expected users of the system.
People or groups who may be impacted.
Sector or programme area.
Data used or processed by the system.
Known risks, safeguards, and mitigation measures.
Stakeholders who should be consulted.
Governance and accountability arrangements.

Areas covered in the EIA template

The EIA template includes questions covering the following assessment areas.

AI system description

Purpose, lifecycle stage, users, impacted people, sector, procurement context, dependencies, and integrations.

Governance

Decision-making authority, diversity of the AI project team, and ethical review responsibilities.

Proportionality and do no harm

Non-AI alternatives, system limitations, human rights, dignity, fundamental freedoms, irreversible impacts, and life-or-death decisions.

Stakeholder engagement

Consultation with potentially impacted stakeholders and mechanisms for taking stakeholder input into account.

Safety and security

Misuse, hacking, user reporting, cybersecurity standards, incident response, testing, revalidation, and data security.

Fairness, non-discrimination and diversity

Bias, representativeness, accessibility, non-discrimination, and diversity-related impacts.

Sustainability

Environmental impact, energy consumption, carbon footprint, monitoring, and public reporting.

Privacy and data protection

Personal data, data minimisation, consent, privacy impact assessment, privacy by design, third-party access, dataset transparency, and ethical sourcing.

Human oversight and determination

Autonomy, human override, escalation to human review, access control, expertise, and safeguards against over-reliance.

Accountability and responsibility

Accountability mechanisms, audits, legal responsibility, response to errors or harms, and claim investigation procedures.

Transparency, explainability and awareness

Public communication, user awareness, access to explanations, understandable outputs, and user guidance.

Practical use in development and humanitarian work

The EIA can support development and humanitarian professionals when AI systems are used or considered in areas such as programme planning, monitoring systems, evidence synthesis, beneficiary feedback analysis, information management, reporting, research, public service delivery, or decision-support processes.

In each case, the EIA helps teams examine whether the AI system has a legitimate purpose, whether affected stakeholders have been considered, whether privacy and fairness risks have been assessed, whether human oversight is clear, and whether accountability mechanisms exist.

Practical questions for development professionals

What is the AI system, and why is it being deployed?
Is the AI system necessary for this purpose?
Were non-AI alternatives explored?
Who will use the system?
Who may be impacted by the system?
Can intended users choose not to use the system?
What personal data does the system process, if any?
How is stakeholder input collected, reviewed, and acted upon?
How could the system affect human rights, dignity, or fundamental freedoms?
How has the system been tested for bias?
Are outputs explainable to non-technical users?
What mechanisms allow human override or human review?
Who is responsible if errors or harms occur?
How will the system be monitored and improved over time?

Connection to monitoring, evaluation, accountability and learning

For monitoring and evaluation professionals, the EIA is relevant because AI systems can influence how information is processed, interpreted, summarised, reported, and used. The EIA helps teams check whether AI use supports evidence quality, stakeholder trust, responsible data use, transparency, and human judgment.

For accountability and learning work, the EIA is useful because it asks teams to identify impacted stakeholders, consider adverse outcomes, document safeguards, and reflect on practical next steps. These are closely connected to responsible evidence use and organisational learning.

Main takeaway

UNESCO’s Ethical Impact Assessment helps teams move from a general commitment to ethical AI toward a structured process for assessing specific AI systems, identifying impacts, engaging stakeholders, and defining mitigation measures.

FAQ

Is the UNESCO Ethical Impact Assessment free?

Yes. UNESCO states that the Ethical Impact Assessment is free to download and use.

Can the EIA be completed asynchronously?

Yes. UNESCO states that the EIA can be completed asynchronously by team members or through one or two half-day collaborative workshops.

Is the EIA only for government?

No. UNESCO identifies several users, including government officials, AI developers and private companies, researchers, and academics.

Should the EIA be used before or after deployment?

UNESCO presents the EIA as a tool for ethics by design and says it is especially valuable early in the AI lifecycle, including research, development and design, procurement, and initial deployment.

Does the EIA include stakeholder engagement?

Yes. UNESCO encourages stakeholder engagement throughout the EIA process, and the EIA template includes questions on consultation with potentially impacted stakeholders.

Why is the EIA relevant for development professionals?

It helps teams examine how AI systems may affect people, data, decisions, accountability, fairness, transparency, and trust in development and humanitarian contexts.

Source

This article is based on UNESCO’s official Ethical Impact Assessment page and the UNESCO EIA template.

Access the UNESCO Ethical Impact Assessment

Note: This article summarises UNESCO’s EIA for educational and professional learning purposes. It does not replace the official UNESCO EIA document.