Explainable Framework Detects Depression Shifts from Digital Traces
A recent study introduces a transparent framework aimed at identifying and examining fluctuations in depression-related states through users' digital footprints, including social media content, conversations, and online engagements. This method utilizes various BERT-based models to gather complementary insights across the dimensions of sentiment, emotion, and depression severity. Over time, these insights are compiled to form user-specific trajectories, which are then scrutinized for significant change points. To improve clarity, the framework incorporates a large language model that produces succinct, easily understandable reports detailing the progression of mental health indicators. This research can be found on arXiv with the identifier 2605.14995.
Key facts
- Framework detects depression status shifts from digital traces
- Uses multiple BERT-based models for sentiment, emotion, and depression severity
- Aggregates signals over time to construct user trajectories
- Identifies meaningful change points in mental health signals
- Integrates a large language model for interpretable reports
- Paper available on arXiv: 2605.14995
- Digital traces include social media posts, chats, online interactions
- Traces are inherently timestamped
Entities
Institutions
- arXiv