DETECTING OF DIGITAL FATIGUE AND BURNOUT IN TEXT CONTENT USING ARTIFICIAL INTELLIGENCE

Abstract

The paper focuses on developing an approach for automated detection of digital fatigue and digital exhaustion in textual content using artificial intelligence methods. The relevance of this study is driven by the growing volume of digital communication in education and professional activities and by the need for scalable, reproducible, and interpretable techniques for monitoring psycho-emotional states without relying on sensor data and without interfering with users’ workflows. The proposed approach is implemented as a sequence of three methods. At the first stage, the author’s digital profile is segmented into communicative segments based on semantic vector representations of texts, followed by clustering and interpretable naming of the segments. At the second stage, each segment is assessed by transformer-based neural models to estimate manifestations of digital fatigue, and segment-level indicators are obtained by aggregating predictions for messages within the segment. A comparison of several architectures is performed, revealing different trade-offs between precision and recall; the most balanced F1 results are achieved by mental/mental-roberta-base and microsoft/deberta-v3-base. At the third stage, a profile-level digital exhaustion indicator is computed as a weighted average aggregation of segment-level fatigue indicators, taking into account the significance of segments within the profile, while interpretability is enhanced by extracting key topics and named entities for each segment. The outputs of the approach are presented as a segment-based digital fatigue map and an integrated digital exhaustion score for the author’s profile. The practical value lies in applying the approach to preventive risk analytics in e-learning systems, remote work environments, and organizational communications, provided that ethical and privacy requirements are observed. Segment-level results can be used for early identification of problematic thematic areas of communication that require workload optimization, adjustments to interaction protocols, or supportive interventions. The integrated digital exhaustion indicator provides a basis for comparing the dynamics of the author’s state over time and for evaluating the effectiveness of implemented organizational measures without disrupting the integrity of work processes.