Impact Of Explainable AI On Cognitive Load: Insights From An Empirical Study

XAI tools often fail end-users despite good intentions: The research reveals that explainable AI systems are typically designed for developers rather than the actual people who need to use them, which helps explain why end-users are frequently reluctant to adopt these systems in practice.

Different types of explanations significantly affect how much mental effort users need: The study of 271 prospective physicians found that the specific way AI explanations are presented strongly influences cognitive load, task performance, and how long tasks take to complete—suggesting that explanation design choices matter substantially for real-world usability.

Local explanations are more mentally efficient than other approaches: Among the explanation types tested in the COVID-19 use case, local XAI explanations (which explain individual predictions rather than overall model behavior) ranked best on a "mental efficiency metric," offering practical guidance for designing more user-friendly AI systems.

Cognitive load is a critical but overlooked factor in XAI design: The paper highlights that previous XAI research has largely ignored how explanations affect users' mental workload, which is essential for understanding whether these systems actually help people make better decisions in real-world settings.

Responsible AI requires interdisciplinary, user-centered research: The findings underscore that developing trustworthy AI systems requires collaboration across fields and empirical testing with actual end-users, not just technical improvements to algorithms.