EUCA: the End-User-Centered Explainable AI Framework

Non-technical explanations are essential but underexplored: The paper identifies that while explainability is critical for AI used in important decisions, most research focuses on technical audiences rather than the everyday people who actually use these systems.

Twelve practical explanation types work for non-experts: The framework identifies twelve explanation formats that don't require technical knowledge—including feature-based (showing which factors mattered), example-based (showing similar past cases), and rule-based explanations (showing decision logic)—providing practitioners with concrete options to choose from.

Different explanation types have different strengths: The user study with 32 layperson participants found that each explanation type has specific pros, cons, and works best for different goals, meaning there's no one-size-fits-all solution; designers must match explanation types to what users actually need to understand.

Co-design with real users reveals actual needs: By involving non-technical participants in prototyping, the research discovered what explanations actually help people make better decisions, rather than assuming what would work—a crucial distinction for responsible AI deployment.

Practical toolkit available for designers and developers: EUCA provides ready-to-use design templates, prototyping cards, and algorithms that AI practitioners can immediately apply to build more understandable AI systems, making responsible explainability more accessible to implement.