Scholarlink

Explainable AI is increasingly embedded in information systems, yet many explanation interfaces remain one-size-fits-all and neglect cognitive heterogeneity and neurodiversity (e.g., ADHD, autism spectrum conditions, dyslexia), which can increase cognitive burden and undermine calibrated reliance. Following Design Science Research, we develop NeuroAdaptX, a neuro-adaptive explanation interface that adapts structure, modality, and density to self-reported profiles while preserving informational equivalence. Grounded in Cognitive Load Theory, Cognitive Fit Theory, and cognitive accessibility guidance, we derive design requirements and consolidate them into design principles for cognitively accessible adaptive explanations. In a randomized between-subjects online experiment (N = 216) against a static baseline, NeuroAdaptX improves objective comprehension, perceived explanation quality, and trust while reducing perceived cognitive load. Overall, the study contributes prescriptive design knowledge for accessibility-oriented neuro-adaptive XAI and empirical evidence that presentation-level adaptation can improve outcomes without changing explanatory substance.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

NeuroAdaptX: Designing Neuro-Adaptive Explanations for Cognitive Accessibility in Explainable AI Interfaces

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Enrico Bunde

摘要

Explainable AI is increasingly embedded in information systems, yet many explanation interfaces remain one-size-fits-all and neglect cognitive heterogeneity and neurodiversity (e.g., ADHD, autism spectrum conditions, dyslexia), which can increase cognitive burden and undermine calibrated reliance. Following Design Science Research, we develop NeuroAdaptX, a neuro-adaptive explanation interface that adapts structure, modality, and density to self-reported profiles while preserving informational equivalence. Grounded in Cognitive Load Theory, Cognitive Fit Theory, and cognitive accessibility guidance, we derive design requirements and consolidate them into design principles for cognitively accessible adaptive explanations. In a randomized between-subjects online experiment (N = 216) against a static baseline, NeuroAdaptX improves objective comprehension, perceived explanation quality, and trust while reducing perceived cognitive load. Overall, the study contributes prescriptive design knowledge for accessibility-oriented neuro-adaptive XAI and empirical evidence that presentation-level adaptation can improve outcomes without changing explanatory substance.