Cognitive distortion based explainable depression detection and analysis technologies for the adolescent internet...

Area of Science:

• Computational Social Science
• Psychology
• Artificial Intelligence

Background:

• Adolescents frequently share daily life on social media, offering data for mental health monitoring.
• Existing depression detection methods primarily use statistical learning, neglecting psychological theories like cognitive distortions.

Purpose of the Study:

• Propose explainable depression detection for adolescents by integrating psychological theory.
• Develop a model that detects depression and identifies cognitive distortions, providing actionable insights.

Main Methods:

• Constructed a multi-task learning model using a pre-trained architecture.
• Trained the model on social media depression and cognitive distortion corpora.
• Employed analytical techniques, including word clouds, for data interpretation.

Main Results:

• The proposed model significantly outperformed baseline methods in depression detection.
• Experimental results demonstrated the model's efficacy in identifying cognitive distortions linked to depression.

Conclusions:

• Social media data can be leveraged for explainable depression detection in adolescents.
• Cognitive distortion theory provides a valuable framework for understanding and detecting adolescent depression.