Evaluating RE Practices for Explainability: Synthesizing Insights from Daimler Truck into an Explainable RE Framework Proposal

Explainability is increasingly critical for safety-critical AI systems, yet it remains difficult to integrate into standard software development lifecycles. This study investigates how practitioners at Daimler Truck apply existing Requirements Engineering (RE) techniques—such as interviews, brainstorming, and compliance checks—to the specific challenge of explainability. The researchers utilized a multi-phase qualitative approach, including think-aloud protocols, to observe how eight practitioners elicited, specified, and validated explainability requirements for an industrial safety feature.

The study identifies a cumulative failure in the RE process. During elicitation, practitioners struggled with a lack of shared terminology, leading to vague and deprioritized requirements. These ambiguities propagated into the specification phase, where techniques like Structured Natural Language and Use Case Diagrams proved insufficient for capturing the cross-cutting, human-centered nature of explainability. Finally, validation was hampered by the lack of measurable acceptance criteria, often reducing the process to linguistic checks rather than meaningful assessments of user understanding or regulatory compliance.

These findings demonstrate that explainability is currently treated as an ad-hoc consideration rather than a first-class requirement. The authors argue that individual RE techniques are not necessarily broken, but rather that the lack of process-level guidance prevents practitioners from maintaining traceability and coherence across the lifecycle. This work serves as a foundation for developing a more robust, empirically grounded RE framework that treats explainability as a systematic, end-to-end concern.