Artificial Intelligence principles and guidelines have proliferated in recent years—transparency, fairness, accountability, and human oversight are widely endorsed. However, when we attempt to implement and operationalise these principles in specific domains, fundamental dilemmas emerge:

In crisis management: How do we balance the need for privacy with transparency when lives are at stake? Can we afford the time for human oversight in rapidly evolving disasters?

In education: How do we ensure fairness in AI-supported learning while respecting pedagogical autonomy and diverse learning needs?

In mobility: When does mandatory human oversight become a safety liability in time-critical traffic situations? How do we operationalize accountability when decisions are distributed across interconnected systems?

This workshop examines a fundamental question for AI policy: What principles have to remain generic across domains, and what should or must be contextualized? The EU AI Act attempts to address this through risk-based categories, but how well does this approach capture domain-specific tensions? Through structured dialogue across domains, we will map where universal principles break down, why contextualisation is necessary, and what this means for developing both sector-specific guidelines and cross-cutting policy frameworks.

This workshop takes places in the “Connect @Mondai”

09.30 – 10.00 Opening & Domain Challenges.
Welcome and short presentations: What are the specific dilemmas when applying AI principles in crisis management, education, and mobility?

10.00 – 10.30 Plenary Principle Mapping.
Interactive session: Starting from the EU Ethics Guidelines for Trustworthy AI and the Dutch Value Compass: Which principles do we prioritize? Where do different domains face irreconcilable conflicts?

10.30 – 11.00 Coffee Break

11.00 – 12.30 Domain Deep Dives, Cross-Domain Synthesis, and Next Steps
> Structured discussions: Develop concrete scenarios where generic principles prove inadequate or create harm, or where additional principles are needed. What makes each domain different? What adaptations are needed?
> Bringing insights together: What patterns emerge? Where is universality possible, and where is contextualization essential?
> Discussion of potential collaborative outputs and future dialogue

Moderator:

• Tina Comes, Scientific Director DLR Institute for the Protection of Terrestrial Infrastructures; Professor in Decision Theory & ICT for Resilience, Delft University of Technology

Expert Speakers:

• Thomas Kox, Head of Research Group “Digitalisation and Networked Security”, Weizenbaum Institute for the Networked Society
• Arkady Zgonnikov, Assistant Professor, Human-Technology Interaction and Centre for Meaningful Human Control, Delft University of Technolog
• Duuk Baten, Advisor on Digitalisation and AI in Education, SURF