Reconfigured boundaries: the dynamics of emotion expression in social-media users reporting a depression diagnosis
Abstract Emotions are multidimensional and dynamic: emotions persist, interact, and change in relation to everyday events. Depression-related differences may therefore emerge both in ongoing emotional dynamics and in responses to particular experiences. Here, we examined both levels using more than 400,000 social-media posts from 1,690 users, including 871 users reporting a depression diagnosis and 819 comparison users without an observed depression-related disclosure. A transformer-based language model inferred normalized probability distributions across 28 emotion-expression dimensions from each post. We then fitted user-specific linear dynamical models using Dynamic Mode Decomposition with Control to characterize post-to-post dynamics and estimate the modeled influence of detected positive life events. Our analyses revealed distinct, emotion-specific patterns of persistence and propagation across the broader emotional landscape. In users reporting a depression diagnosis, these ongoing dynamics were fundamentally reconfigured: the distinct roles of several emotions blurred, while others shifted in opposing directions. This points to a reorganization of emotional boundaries rather than a simple increase in negativity or a uniform flattening of affect. In contrast, immediate responses to positive life events remained largely preserved after correcting for post count. Together, these findings suggest that depression-related differences manifest primarily as a structural reconfiguration of baseline emotional dynamics rather than a dampened capacity to respond to positive events.
Authors
- Sharmili Edwin Thanarajah (ORCID: https://orcid.org/0000-0003-3608-1474)
- Ana-Maria Bucur (ORCID: https://orcid.org/0000-0003-2433-8877)
- Hamidreza Jamalabadi (ORCID: https://orcid.org/0000-0003-2485-2374)
- Paolo Rosso (ORCID: https://orcid.org/0000-0002-8922-1242)
- Felix P. Bernhard
- Adrian Cosma (ORCID: https://orcid.org/0000-0003-0307-2520)
- Lucie Flek (ORCID: https://orcid.org/0000-0002-5995-8454)
- Bita Shariatpanahi
- Tahmineh A. Koosha
Institutions
- Goethe University Frankfurt (DE)
- University of Bonn (DE)
- University of Bucharest (RO)
- Philipps University of Marburg (DE)
- Dalle Molle Institute for Artificial Intelligence Research (CH)
- Max Planck Institute for Metabolism Research (DE)
- University Hospital Frankfurt (DE)
- Bonn Aachen International Center for Information Technology (DE)
- Lamarr Institute for Machine Learning and Artificial Intelligence (DE)
- Università della Svizzera italiana (CH)
- Universitat Politècnica de València (ES)
Publication Details
- Journal
- Scientific Reports
- Published
- 2026-11-17
- DOI
- https://doi.org/10.1038/s41598-026-71638-4
- Primary Topic
- Mental Health via Writing
- Type
- article
- Field-Weighted Citation Impact
- 0.00
Funders
- Deutsche Forschungsgemeinschaft
- Bundesministerium für Bildung und Forschung
- Uniklinikum Giessen und Marburg
- European Regional Development Fund
- Agencia Estatal de Investigación