Proceedings of the International scientific and practical conference ―Synergy of Modern Science and Education‖ (February 2-4, 2026) / Publisher website: www.naukainfo.com. – New York, USA, 2026. - 324 p.

73  Long pauses, hesitations before confirming an action. Combined patterns are especially important: for example, the combination of increased time per step and a higher error frequency often signals cognitive overload more reliably than individual indicators. It was shown that detecting confusion from the sequence of user actions achieves accuracy of 72–78%, which is practically acceptable for educational systems [12]. Taken together, these patterns can be sufficient for an ―overload risk signal,‖ even if eye data are unavailable. To validate the detector, one can add a short question after a module: for example, 9-point mental effort scale (―from very, very low to very, very high‖). This is one of the classic subjective measures of cognitive load and is easy to integrate into the learning flow [13]. For practical implementation, a simple discrete model is sufficient:  Green (OK): indicators are stable, errors are within the norm (≤ 2 per task), pace is even (deviation < 20% from baseline);  Yellow (Mild overload): increase in time per step > 30% from baseline + > 2 errors per task; backtracking appears (> 2 returns per module) or unstable eye patterns (increase in fixation duration > 40%, decrease in blink rate);  Red (Overload/fatigue): slowing > 50% + high error rate (> 3 attempts per task) or refusal of 2+ tasks in a row, systematic long pauses (> 60 seconds without activity), or critical changes in ocular metrics. The key idea is not to measure an ―emotion,‖ but to detect a functional state that reduces learning effectiveness and apply an interface intervention. Here, progressive disclosure (PD) could become a mechanism for reducing extraneous load. PD is a principle whereby secondary/advanced elements and information are revealed gradually, while basic actions remain on the surface. This reduces interface complexity, increases clarity, and lowers errors, especially for novices or in a state of overload [14]. In learning, progressive disclosure can be applied at two levels. One of them is Interface (UI), where the following measures could be implemented: