15 - Guest lectures
EXPLAINABILITY AT ALL LEVELS OF SYSTEM DEVELOPMENT: INSIGHTS FROM THE DESIGN OF INTELLIGENT SYSTEMS IN AVIATION
J. Würfel¹, B. Djartov¹, M. Friedrich¹, S. Schog¹, A. Papenfuss¹; ¹German Aerospace Center (DLR), Germany
This paper proposes a multi-level approach to operational explainability in aviation AI systems, combining bottom-up model design with top-down human-centered interface development. Using civil and military use cases, a simulator study shows that combining factor ratings with feature contributions improves perceived explainability and decision support effectiveness.
DOI: 10.71945/icas2026_0177