The Role of Normware in Trustworthy and Explainable AI

Normware is a distinct category from software that deserves separate attention: The paper argues that computational artifacts specifying norms (normware) cannot be reduced to or treated the same as traditional software, and require their own conceptual framework for responsible AI development.

Current AI systems lack sufficient explainability and comprehensibility: The paper identifies that contemporary technology is "incomprehensible" and "unintelligible" to most people, creating a significant challenge for trustworthiness that existing approaches have not adequately addressed.

The "ecological dimension" of AI decision-making has been neglected: The paper emphasizes that normware's role in the decision-making cycle includes broader ecological and societal impacts that are typically overlooked in technical discussions, suggesting a more holistic approach is needed.

New conception methods are urgently needed for AI governance: Rather than relying solely on existing technical or policy approaches, the paper calls for fundamentally reorganized ideas and discussion methods to address how norms are embedded in AI systems.

Trustworthiness requires explicit attention to the norms embedded in AI: By making normware visible and explicit rather than hidden within code, organizations can better ensure their AI systems operate according to clearly defined values and principles.