Exploring Mental Illness Stigma, Help-Seeking Attitudes, and Behaviors on Social Media: A Machine Learning Approach

Introduction Mental health literacy (MHL) plays a crucial role in promoting help-seeking behavior. However, negative attitudes toward mental illness still pose a substantial barrier. Social media platforms provide a valuable opportunity to explore the relationship between stigma, help-seeking attitudes, and behaviors through the application of natural language processing (NLP) and machine learning (ML) techniques. Objectives This study aims to investigate how attitudes toward reducing mental illness stigma and help-seeking influence actual help-seeking behavior among social media users. Methods We analyzed 1,506,333 posts from mental health-related Post Text Table (PTT) forums between January 2018 and January 2024. Posts were preprocessed and categorized into three categories: reducing mental illness stigma, help-seeking attitudes, and help-seeking behaviors. Using Bidirectional Encoder Representations from Transformers (BERT) for scoring, the model achieved a precision of 0.81 and accuracy of 0.89. Logistic regression was then applied to assess the predictive value of stigma reduction and help-seeking attitudes on help-seeking behavior. Results The study found that for each one-unit increase in score measuring attitudes toward reducing mental illness stigma, he likelihood of help-seeking behavior increased by 1.35 times (95% CI: 1.21–1.50, p < 0.001). Similarly, stronger help-seeking attitudes were associated with a 1.41 times higher likelihood of help-seeking behavior (95% CI: 1.16–1.71, p < 0.05). Binary logistic regression analysis further demonstrated that users with more pronounced stigma-reducing attitudes and positive help-seeking attitudes were 1.76 times (95% CI: 1.43–2.17, p < 0.001) and 2.62 times (95% CI: 1.48–4.65, p < 0.05) more likely to engage in help-seeking behavior, respectively. Conclusions This study highlights that stronger attitudes toward reducing mental illness stigma, along with more positive help-seeking attitudes, significantly predict help-seeking behavior. By leveraging machine learning and natural language processing, it offers novel insights into how social media discussions influence mental health behaviors, providing a valuable foundation for future interventions aimed at reducing stigma, fostering positive attitudes toward help-seeking, and encouraging actual help-seeking behaviors. Disclosure of Interest None Declared