Explainable software systems: from requirements analysis to system evaluation

Abstract The growing complexity of software systems and the influence software-supported decisions in our society sparked need for that is transparent, accountable, trustworthy. Explainability has been identified as a means to achieve these qualities. It recognized an emerging non-functional requirement (NFR) significant impact on system quality. Accordingly, engineers assist them incorporating this NFR into systems. This requires early analysis benefits possible design issues arise from interrelationships between different quality aspects. However, explainability currently under-researched domain requirements engineering, there lack artifacts support engineering process design. In work, we remedy deficit by proposing four artifacts: definition explainability, conceptual model, knowledge catalogue, reference model explainable These should understanding how it interacts with other Besides that, they may be considered starting point provide practical value refinement high-level concrete choices, well identification methods metrics evaluation implemented requirements.