Beyond Technocratic XAI: The Who, What & How in Explanation Design

Explanation design requires intentional choices beyond just technical methods: The paper argues that simply applying XAI techniques isn't enough—practitioners must make deliberate design decisions about who receives explanations, what information they need, and how it should be presented to ensure the explanations are actually meaningful and accessible.

Different stakeholders need different explanations: The framework emphasizes asking "Who needs the explanation" because regulators, end-users, and developers have fundamentally different needs. A one-size-fits-all approach to AI transparency fails to serve any group effectively.

Poorly designed explanations can reinforce existing inequalities: The paper warns that explanations risk creating "epistemic inequality"—where some groups gain understanding while others are left behind—and can inadvertently reinforce social inequities if not carefully designed with diverse audiences in mind.

Explanations can obscure rather than clarify accountability: The authors caution that explanations might actually hide who is responsible for AI decisions rather than illuminate it, potentially undermining governance and accountability if explanation design doesn't explicitly address these concerns.

Explanation is a design problem, not just a technical problem: By framing XAI as a "sociotechnical design process," the paper suggests that building trustworthy AI requires input from designers, ethicists, and domain experts—not just computer scientists—to ensure explanations serve their intended purpose responsibly.