Modelling GDPR-Compliant Explanations for Trustworthy AI

GDPR's Right to Explanation requires companies to provide understandable explanations for automated decisions, and this paper proposes a practical framework called "Explanatory Narratives" to help organizations meet this legal requirement while making explanations genuinely useful to people affected by AI decisions.

One-size-fits-all explanations don't work: The paper emphasizes that different people need different types of explanations depending on their background and goals, so effective AI systems should allow users to interactively explore explanations and customize them to their specific needs rather than receiving a single fixed explanation.

The framework treats explanation-generation as a guided exploration process where users can navigate through different types of information about how an AI decision was made, similar to how someone might explore a map by choosing different routes—this interactive approach aligns with both GDPR requirements and the EU's AI-HLEG guidelines for human-centered AI.

The paper identifies key design principles from ISO 9241 (usability standards) that should guide how explanations are presented, suggesting that responsible AI requires applying established human-centered design practices rather than treating explanations as purely technical outputs.

The authors acknowledge both strengths and limitations of their model, demonstrating intellectual honesty about the challenges in creating truly compliant and user-friendly explanations—an important consideration for organizations implementing these concepts in practice.