Foundations of Explainable Knowledge-Enabled Systems

Explainability has deep roots in AI history: The paper traces explainability back to the early days of artificial intelligence, showing this isn't a new concern but rather a foundational goal that has evolved alongside AI technology itself.

Past AI approaches offer valuable lessons: Expert systems and cognitive assistants from earlier eras developed explanation methods that remain relevant today, suggesting that responsible AI development can benefit from revisiting and adapting proven approaches rather than starting from scratch.

Critical settings demand better explanations: The paper emphasizes that as AI systems are increasingly deployed in high-stakes environments, the ability to explain decisions to end-users and decision-makers has become essential—not optional—for responsible deployment.

One-size-fits-all explanations don't work: The authors identify a key gap in existing approaches: explanations must be tailored to specific users and contexts rather than being generic outputs, meaning responsible AI requires understanding who needs to understand the system and why.

New frameworks are needed for modern systems: The paper proposes updated definitions for explainability and explainable systems that address limitations of past approaches, suggesting that responsible AI requires continuously refining our understanding of what "explainability" actually means in practice.