A psycho-ecological signal recognition framework for user behavior prediction on digital media platforms

Area of Science:

• Digital behavioral science
• Computational psychology
• Human-computer interaction

Background:

• Digital media is pervasive, but excessive or emotionally charged use, particularly at night, raises behavioral and mental health concerns.
• Current predictive models overlook the complex interaction between users' psychological states and their environmental contexts.
• Understanding these dynamics is crucial for developing effective digital wellness strategies.

Purpose of the Study:

• To develop a novel behavior prediction model integrating psychological and ecological factors for digital media usage.
• To identify high-risk patterns associated with immersive use, late-night activity, and susceptibility to misinformation.
• To lay the groundwork for early-stage intervention strategies in digital health.

Main Methods:

• Proposed a Dual-Channel Cross-Attention Network (DCCAN) with signal identification, interaction modeling, and behavior prediction layers.
• Employed cross-modal attention mechanisms to fuse psychological and ecological data.
• Trained and validated the model on a large dataset of 9,782 users and 51,264 behavior sequences.

Main Results:

• The DCCAN model significantly outperformed baseline models (LSTM, GRU, XGBoost) in predicting immersive usage, late-night activity, and misinformation susceptibility.
• Achieved high performance metrics, including an F1-score of 0.891 and AUC of 0.913 for immersive usage prediction.
• Ablation studies confirmed the importance of psychological/ecological signals and the cross-attention mechanism.

Conclusions:

• Integrating psychological and ecological data via attention-based fusion provides accurate and interpretable predictions of digital risk behaviors.
• The DCCAN framework offers a promising approach for real-time behavioral health monitoring and adaptive content moderation.
• This research advances the understanding of digital media's impact on mental well-being and intervention possibilities.