Charting the Sociotechnical Gap in Explainable AI: A Framework to Address the Gap in XAI

The "sociotechnical gap" is a critical but often overlooked problem in XAI: The paper identifies that explainable AI systems fail not because the technology is flawed, but because there's a disconnect between what the technical systems can provide and what people actually need in real-world contexts. This gap exists in distinct domains and requires systematic attention.

Technical explanations alone are insufficient for responsible AI: The research demonstrates through case studies that even well-designed technical explanations may not meet users' actual needs. This suggests that organizations implementing XAI systems must go beyond simply adding explanation features and instead understand the specific social context where the AI will be used.

A framework-based approach can help organizations identify and address XAI gaps: The paper proposes a practical framework that connects AI guidelines to real-world needs, allowing organizations to systematically chart where their explainability efforts fall short. This provides actionable steps rather than abstract principles for improving AI transparency.

Different domains require different explainability solutions: By testing their framework across multiple domains (two initial case studies plus a third validation case), the authors show that there is no one-size-fits-all approach to explainable AI. Organizations must tailor their explainability strategies to their specific context and user needs.

Charting the gap improves problem-solving: The paper argues that explicitly mapping the sociotechnical gap—rather than ignoring it—leads to better solutions. This reflexive approach helps organizations design more responsible AI systems that actually serve their intended users effectively.