Orchestrating Attention: Bringing Harmony to the 'Chaos' of Neurodivergent Learning States

TL;DR

The paper tackles dynamic attention fluctuations in neurodivergent learning, specifically ADHD, by introducing AttentionGuard, a framework that detects four engagement-attention states using privacy-preserving behavioral signals and adapts UI through five state-responsive patterns. The approach combines a 30-second-window feature set with a Random Forest classifier to produce four states (Focused, Drifting, Hyperfocused, Fatigued) and applies bi-directional UI adaptations designed to support both overstimulation and understimulation. Empirical validation includes an 87.3% detection accuracy on the OULAD dataset, cross-dataset correlation with ADHD profiles (AUC 0.81; r=0.47), and a Wizard-of-Oz pilot (N=11) showing substantial reductions in cognitive load (d=1.21) and improved comprehension (d=1.18), with high concordance between automated predictions and human decisions (κ=0.71). The work emphasizes interface-level transparency and user agency, providing a concrete design vocabulary of patterns—plus a demo mode enabling inspection and contesting of AI-driven adaptations—demonstrating practical potential for neurodiversity-aware intelligent interfaces.

Abstract

Adaptive learning systems optimize content delivery based on performance metrics but ignore the dynamic attention fluctuations that characterize neurodivergent learners. We present AttentionGuard, a framework that detects engagement-attention states from privacy-preserving behavioral signals and adapts interface elements accordingly. Our approach models four attention states derived from ADHD phenomenology and implements five novel UI adaptation patterns including bi-directional scaffolding that responds to both understimulation and overstimulation. We validate our detection model on the OULAD dataset, achieving 87.3% classification accuracy, and demonstrate correlation with clinical ADHD profiles through cross-validation on the HYPERAKTIV dataset. A Wizard-of-Oz study with 11 adults showing ADHD characteristics found significantly reduced cognitive load in the adaptive condition (NASA-TLX: 47.2 vs 62.8, Cohen's d=1.21, p=0.008) and improved comprehension (78.4% vs 61.2%, p=0.009). Concordance analysis showed 84% agreement between wizard decisions and automated classifier predictions, supporting deployment feasibility. The system is presented as an interactive demo where observers can inspect detected attention states, observe real-time UI adaptations, and compare automated decisions with human-in-the-loop overrides. We contribute empirically validated UI patterns for attention-adaptive interfaces and evidence that behavioral attention detection can meaningfully support neurodivergent learning experiences.

Figures (4)

• Figure 1: Architecture of the AttentionGuard Framework. Privacy-preserving behavioral signals (interaction, response, and navigation) are aggregated over 30-second windows and compared to a personalized baseline (5-minute initial calibration). A Random Forest classifier detects one of four ADHD-derived attention states: Focused (productive engagement), Drifting (attention wandering), Hyperfocused (deep absorption), and Fatigued (depleted capacity). The Adaptation Engine applies bi-directional modulation and triggers five neuroscience-grounded UI patterns. Adaptations are visible and reversible to support interface-level transparency, with Wizard-of-Oz concordance (84% exact match, $\kappa$=0.71) validating automated feasibility. A continuous personalization loop sustains improved comprehension and reduced cognitive load.
• Figure 2: The Companion System (S1). A sidebar element providing Virtual Body Doubling to reduce isolation paralysis and increase task salience.
• Figure 3: Your Thinking Space (S2). A tool allowing users to externalize spontaneous thoughts via text or voice without leaving the reading flow.
• Figure 4: Interest Injection System (S3). A gamified module delivering context-aware curiosity hooks and micro-rewards (+XP) to sustain dopamine levels.