AI
New AI Model Detects Cognitive Distortions in Text for Mental Health Support
Researchers have developed an interpretable machine learning approach to automatically identify patterns of distorted thinking in natural language, potentially helping psychologists and therapists screen patient communications at scale.
A research team has unveiled a new computational method for identifying cognitive distortions—unhelpful thought patterns common in mental health conditions—directly within written text. The approach uses weighted structured patterns and N-grams (short word sequences) to classify and detect these distortions, prioritizing transparency and explainability in its decision-making.
The Method#
The system relies on interpretable artificial intelligence rather than opaque deep learning models, making it possible for clinicians to understand exactly why the system flagged particular passages. The algorithm accounts for complex relationships between linguistic patterns, treating language as a hierarchical system where some features matter more than others.
Performance Gains#
When tested on two publicly available datasets, the new approach substantially outperformed existing methods, achieving higher F1 scores—a standard measure of classification accuracy that balances precision and recall. The researchers optimized the model's hyperparameters for real-world use and have made their code and trained models available to the research community for further development.
Clinical Potential#
The work addresses a pressing need in psychological care: automating the detection of specific distorted thinking patterns that individuals may express in their own words. By doing so, it could help therapists prioritize cases, monitor patient progress, or flag concerning thought patterns during digital mental health interventions.
