Meaningful XAI Based on User-Centric Design Methodology

XAI serves five distinct legal purposes: The paper identifies that explainability requirements across EU regulations aren't one-size-fits-all, but rather serve specific purposes including contestability (the ability to challenge decisions), empowerment (reducing information gaps between users and systems), performance control, decision evaluation, and transparency in public administration.

Different stakeholders need different explanations: The paper identifies four key recipient groups—data scientists, human operators, affected individuals, and regulators—each requiring tailored explanations rather than generic technical documentation, emphasizing that meaningful XAI must match the audience's needs and expertise level.

Explainability must be designed for specific operational moments: Rather than treating XAI as a single feature, the paper identifies four critical contexts where explanations matter: during system design and testing, during real-time human oversight, during human decision-making with AI input, and when investigating or challenging decisions after deployment—each requiring different explanation approaches.

User-centered design is essential for effective AI transparency: The paper argues that meaningful XAI cannot be achieved through technical solutions alone; instead, it requires deliberate design methodology that considers the specific purpose, intended audience, and operational context to ensure explanations are actually useful rather than merely technically accurate.

One-size-fits-all explanations undermine responsible AI goals: The research suggests that generic explainability features often fail to serve their intended legal and ethical purposes, highlighting the need for customized XAI solutions developed through stakeholder engagement rather than standardized technical implementations.