Identifying Explanation Needs of End-users: Applying and Extending the XAI Question Bank

Expert users ask questions beyond what current XAI frameworks capture: The study found that subject matter experts raised explanation needs not covered by the existing XAI Question Bank, suggesting that widely-used frameworks for identifying explanation needs may be incomplete in real-world usage contexts.

Generic AI explanations fail because they ignore user expertise and context: The paper confirms that explanations designed by AI experts focusing on algorithmic transparency don't work for actual users, highlighting the critical gap between how AI systems are explained and what different users actually need to understand.

The XAI Question Bank, despite being influential, lacks practical validation: While the XAIQB is widely referenced in the field, the authors found few reports of its actual use in practice, revealing a disconnect between theoretical frameworks and their real-world application in identifying what users need explained.

A human-centered approach requires understanding specific usage contexts: The research demonstrates that effective explanations cannot be one-size-fits-all; instead, developers must investigate the particular explanation needs of end-users (both subject matter experts and lay users) within their specific contexts of use.

Current XAI tools need refinement to serve practitioners: The study's findings about missing explanation needs suggest that frameworks guiding XAI development should be updated and extended based on empirical evidence from actual user interactions, not just theoretical considerations.