Understanding User Preferences in Explainable Artificial Intelligence: A Survey and a Mapping Function Proposal

One-size-fits-all explanations don't work: The paper identifies a critical gap in XAI research—while demand for explainable AI is growing, most studies fail to match different explanation methods to individual user needs, meaning current approaches may not actually serve the people who need to understand AI systems.

Users need personalized explanation methods: The researchers propose a "mapping function" that matches specific users and their needs to appropriate XAI methods, recognizing that a doctor, regulator, and data scientist likely need different types of explanations from the same AI system.

XAI methods vary significantly in their properties: The study categorizes explainability approaches across three domains—philosophy, theory, and practice—suggesting that understanding these distinctions is essential for practitioners to select methods that actually address their specific use case rather than adopting explanations arbitrarily.

Tailoring explanations improves practical value: By connecting users' desired properties (such as transparency, trustworthiness, or fairness) to specific XAI methods, the research suggests that customized explanations are more likely to achieve meaningful goals in responsible AI deployment.

Bridging the research-practice gap is urgent: The paper highlights that despite extensive XAI research, there remains a scarcity of practical guidance on which methods work best for whom, indicating that responsible AI implementation requires better translation of academic work into actionable tools for real-world users.