As Artificial Intelligence becomes increasingly relevant to labour market analysis and policymaking, its development must be guided by fairness, transparency, accountability and respect for fundamental rights.
To strengthen this approach, the EU-ALMPO consortium participated in two complementary Ethics Workshops led by Dr Tomasz Hollanek, Assistant Research Professor at the Leverhulme Centre for the Future of Intelligence, University of Cambridge. Together, the workshops combined theoretical foundations, regulatory guidance and practical exercises connected directly to the AI-powered tools being developed within the project.
The workshops introduced participants to the relationship between AI ethics and AI governance. Discussions covered key principles of responsible AI, including privacy, safety, transparency, explainability, fairness, non-discrimination, human oversight and professional responsibility. Participants also reflected on how technologies influence values and decision-making and considered whose interests and perspectives are represented when an AI system is designed.
A central message was that ethical AI cannot be reduced to a checklist or treated as a final compliance step. Ethical questions depend on the specific context in which a system is developed and used, while different groups may experience the same technology in very different ways. Responsible AI therefore requires continuous reflection, informed judgement and governance throughout the system’s lifecycle.
The workshops also examined the regulatory and institutional mechanisms supporting trustworthy AI, from international guidance and organisational self-governance to binding legislation. Particular attention was given to the EU AI Act, its risk-based classification system and the responsibilities of AI providers and deployers. This is especially important for systems applied in employment, vocational training and access to essential public services, which may fall under the Act’s high-risk categories.
These discussions were then connected directly to the EU-ALMPO project. EU-ALMPO is grounded in the European values of fairness, dignity and inclusion and is guided by the EU AI Act, the General Data Protection Regulation and the EU Charter of Fundamental Rights. Its ethical approach focuses on non-discrimination, accountability, stakeholder co-design, human oversight and adaptation across different national labour market systems.
Participants identified several areas requiring continuous attention, including algorithmic bias, privacy and data protection, data quality, the use of third-party datasets, stakeholder participation and the wider societal effects of AI-supported labour market decision-making.
Moving from principles to practice, participants explored how to identify everyone who may be affected by an AI system. This included not only direct users, but also people evaluated by the system, employment professionals interpreting its recommendations, system providers, policymakers, civil society organisations and communities that may be indirectly affected.
The exercises emphasised the importance of intersectional and inclusive design. Factors such as age, gender, disability, ethnicity, educational background and family circumstances may interact and influence how people experience an AI system. Participants were therefore encouraged to question the idea of a single “typical user” and consider who is represented, who may be excluded and who is involved in the design process.
The workshops also introduced principles of Data Justice and Data Feminism, highlighting that data is not neutral. Its collection, classification and interpretation reflect particular social contexts, assumptions and power relations. Strong data governance therefore requires transparency about where data comes from, how it was created, whose experiences it represents and where important gaps or biases may exist.
Another central topic was stakeholder participation. Participants distinguished between one-off consultation and meaningful participation throughout the development lifecycle. Responsible AI requires affected groups to have a genuine opportunity to influence the purpose, design, implementation and evaluation of a system, rather than simply being asked to provide feedback after key decisions have already been made.
The EU-ALMPO AI-powered CV Recommendation Tool was used as a practical case study. The tool is intended to analyse structured CV data alongside labour market intelligence, recommend suitable occupations, training opportunities and Active Labour Market Policy interventions, and identify possible upskilling or reskilling pathways.
Through Consequence Scanning, participants examined both the expected benefits and potential unintended effects of the tool. Possible risks included excessive reliance on automated recommendations, the exclusion of atypical career paths, reinforcement of historical labour market inequalities and reduced autonomy for employment advisers.
Participants considered how these consequences might develop after six months, two years or wider adoption across Europe. They were also asked to identify which values and fundamental rights could be affected, who would benefit, who might experience additional burdens and whether the proposed safeguards would be sufficient.
The two workshops ultimately highlighted four priorities for EU-ALMPO: embedding fundamental rights, fairness and accessibility into system design; expanding ethical risk assessment beyond minimum legal compliance; strengthening data governance and documentation; and ensuring that stakeholder participation is meaningful, continuous and sensitive to power differences.
For EU-ALMPO, ethical AI is not an abstract principle. It is an ongoing practical responsibility that must shape how AI systems are designed, tested, governed and used.