Textual forma mentis networks bridge language structure, emotional content and psychopathology levels in adolescents

AI can detect mental health patterns in how adolescents use language: Researchers developed a system that analyzes the structure and emotional content of spoken words to predict psychological difficulties, finding meaningful correlations with social maladjustment (r = 0.37), internalizing behaviors (r = 0.33), and neurodevelopmental risk (r = 0.34) in 232 adolescents.

The way we organize thoughts linguistically may reflect psychological struggles: Adolescents showing signs of social maladjustment displayed more "clustered" language patterns (higher modularity and core-periphery structure), suggesting their conceptual thinking is more compartmentalized, while those with internalizing behaviors showed linguistic markers associated with rumination, like higher emotional disgust expressions.

This approach combines multiple dimensions of language for better insights: Rather than analyzing words in isolation, the framework examines how words connect syntactically (grammar), semantically (meaning), and emotionally, providing a more complete picture of psychological state than traditional text analysis methods.

Explainability matters for clinical applications: The researchers used SHAP analysis to identify which specific language features predicted mental health outcomes, moving beyond "black box" predictions to show clinicians and patients what linguistic patterns are most meaningful—essential for responsible deployment in healthcare settings.

This is early-stage research requiring validation: The study used a single dataset of structured interviews on one emotion-inducing topic; broader testing across diverse populations, conversation types, and clinical settings would be needed before this could inform real clinical decisions.