AI Research Explores Tracking Mental Health Changes in Social Media Timelines with Privacy Focus
Research presented by MSc Computer Science student Maryia Zhyrko at the Leiden AI Community focuses on using Artificial Intelligence (AI) to identify changes in mental health signals within social media timelines. The project, named DreamerNLplus, employs a hybrid approach combining local language models, rule-based methods, and traditional machine learning. This framework aims to analyze social media data to detect evolving psychological states over time while prioritizing user privacy by keeping sensitive information under local control. The research moves beyond simple sentiment analysis, striving to recognize deeper psychological states and their development across a sequence of posts. A central objective is interpretability, ensuring that clinicians and researchers can understand how AI predictions are reached, addressing concerns about reproducibility and the 'black box' nature of some AI models.
