What Survives Legalisation? Governance, Enforceability, and the Filtering of Stakeholder Beliefs in the EU AI Act

This thesis examines how stakeholder beliefs expressed during the European Commission’s 2020 consultation on artificial intelligence relate to the content of the EU Artificial Intelligence Act adopted in 2024. Drawing on the Advocacy Coalition Framework, it analyses consultation submissions as structured expressions of competing problem definitions rather than isolated organisational positions. The analysis finds that disagreement in EU AI governance clusters around coherent belief configurations cutting across industry, civil society, and public authorities. As policy moves from agenda-setting to binding law, beliefs articulated in governance-operable terms such as regulatory design, oversight, and administrative responsibility, persist more visibly in legislative text than broad normative or strategic claims. This pattern reflects the institutional demands of EU legalisation, which privilege enforceable obligations and administratively workable procedures. The findings support...Show moreThis thesis examines how stakeholder beliefs expressed during the European Commission’s 2020 consultation on artificial intelligence relate to the content of the EU Artificial Intelligence Act adopted in 2024. Drawing on the Advocacy Coalition Framework, it analyses consultation submissions as structured expressions of competing problem definitions rather than isolated organisational positions. The analysis finds that disagreement in EU AI governance clusters around coherent belief configurations cutting across industry, civil society, and public authorities. As policy moves from agenda-setting to binding law, beliefs articulated in governance-operable terms such as regulatory design, oversight, and administrative responsibility, persist more visibly in legislative text than broad normative or strategic claims. This pattern reflects the institutional demands of EU legalisation, which privilege enforceable obligations and administratively workable procedures. The findings support institutional filtering as a plausible explanation for why extensive consultation yields selective legislative outcomes, raising broader questions about how participation translates to representation in EU technology governance.Show less

Leiden University Student Repository

Documents

In Collections

What Survives Legalisation? Governance, Enforceability, and the Filtering of Stakeholder Beliefs in the EU AI Act