A deep learning and causal inference framework for analysing mental health discourse on social media networks
This paper examines how deep learning and causal inference techniques are combined to analyse mental health coping stories posted on social media sites. Individuals often reveal personal experiences linked to mental health issues and coping strategies as user-generated content on sites like Twitter,...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Conference or Workshop Item |
| Language: | en |
| Published: |
IEEE
2026
|
| Subjects: | |
| Online Access: | https://umpir.ump.edu.my/id/eprint/47471/1/A%20deep%20learning%20and%20causal%20inference%20framework.pdf https://umpir.ump.edu.my/id/eprint/47471/ https://doi.org/10.1109/ICICSE67247.2025.11390793 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | This paper examines how deep learning and causal inference techniques are combined to analyse mental health coping stories posted on social media sites. Individuals often reveal personal experiences linked to mental health issues and coping strategies as user-generated content on sites like Twitter, Facebook, and Reddit grows quickly. Traditional observational studies struggle to establish causal links because of confounding variables and biases inherent in social media data. Among recent developments in deep learning, natural language processing (NLP) models provide powerful tools for processing and interpreting vast amounts of unstructured text data. Causal inference methods, on the other hand, provide frameworks that help one deduce cause-and-effect links, even in the face of the difficulties that observational data presents. This paper synthesizes existing research by combining these approaches to examine their use in understanding the mental health coping mechanisms of online communities. It also discusses ethical issues, model interpretability, and data quality questions. This article presents a systematic review and framework that combines deep learning and causal inference methods to study mental health discourse on social media. The integration of these approaches addresses biases in observational data and enhances the causal understanding of coping narratives. Dominant trends and key models, such as BERT and LSTM, are identified, along with a discussion of the ethical and methodological challenges. Lastly, the study outlines future research paths, including building explainable AI models, conducting longitudinal studies, analysing cross-platform data, and developing ethical frameworks to guide responsible research behaviour. |
|---|