Model and algorithm for neural network-based gaze fixation duration determination for dynamic content adaptation in inclusive learning

The relevance of this study stems from the fact that most existing inclusive educational technologies provide only basic access to content but do not ensure its effective mastery. In practice, they rely on static personalization grounded in a pre-defined user profile (e.g., impairment type). The key shortcoming of this approach is its blindness to the dynamics of the actual learning process: the system cannot recognize when a learner experiences an immediate spike in cognitive load or encounters difficulty understanding a specific term or formula. This creates a barrier that leads to frustration, loss of motivation, and ultimately superficial learning.The objective of the work is to develop and substantiate an innovative mechanism for real-time dynamic adaptation of educational content that responds to learners’ cognitive difficulties as they occur.The study is based on designing a system that uses eye-tracking to monitor user interaction with digital instructional materials. The primary biometric indicator is the fixation duration on specific semantic units (text fragments, images, formulas).This work is the first to propose an adaptation mechanism grounded in the hypothesis that prolonged gaze fixation is a reliable marker of comprehension difficulty or elevated cognitive load. Upon detection of an extended fixation, the system automatically triggers an adaptive scenario and offers targeted, context-specific support (e.g., term definitions, simplified explanations, visual examples) precisely for the unit that elicited difficulty. This approach enables a shift from macro-personalization (profile-level) to micro-personalization (real-time), thereby creating a more responsive and effective inclusive learning environment.