Explainable AI does not provide the explanations end-users are asking for

XAI tools often fail to meet user needs: The paper finds that explainable AI techniques frequently don't provide the types of explanations that end-users actually want, suggesting a significant gap between what XAI developers create and what stakeholders require to understand AI systems.

XAI has unintended consequences when deployed: While XAI methods may be useful during development and testing phases, their real-world adoption by organizations can backfire, potentially creating false confidence or misleading users rather than genuinely enhancing trust.

Transparency and validation are more effective than XAI alone: The authors argue that rigorous validation of AI systems combined with transparent documentation about how they work and their limitations provides a better foundation for trustworthy AI than relying on XAI explanations.

Trust requires more than technical explanations: The research suggests that building genuine trust in AI systems involves more than making models technically explainable—it requires comprehensive approaches that address user concerns and organizational accountability.