In this work, Marta Witkowska and colleagues examine whether subtle patterns of semantic agency in everyday language can reveal underlying psychological distress. Across two studies, the research team analyzed large-scale social media data from Twitter/X and Reddit, combining machine learning-based topic detection with established linguistic markers and expert coding.
Their findings reveal that posts related to depressive experiences—particularly during the postpartum period—consistently exhibit lower levels of semantic agency. This suggests that a reduced expression of control, action, and goal-directedness in text can serve as a meaningful digital marker for psychological distress. By highlighting how the subtle nuances of daily online communication can help identify people in need of support, this work contributes to ongoing research in social cognition, computational linguistics, and digital mental health. It will be of interest to researchers, clinicians, and anyone working at the intersection of language and psychological well-being.
Publication:
Witkowska, M., Beneda, M., Formanowicz, M., Arslan, S., Nikadon, J., Kowalski J., Erseghe, T., & Suitner, C. (2026). The Semantics of depression: How linguistic agency patterns signal depressive symptoms on social media. Advance online publication. Depression & Anxiety. https://doi.org/10.1155/da/6485997
Referenced projects:
Keywords: agency; depression; linguistic markers; postpartum; social media