Computational modeling of stigmatized behaviour in pro-vaccination and anti-vaccination discussions on social media

Much research has been done within the social sciences on the interpretation and influence of stigma on human behaviour and health, which result in out-of-group exclusion, distancing, cognitive separation, status loss, discrimination, in-group pressure, and often lead to disengagement, non-adherence to treatment plan, and prescriptions by the doctor. However, little work has been conducted on computational identification of stigma in general and in social media discourse in particular. In this paper, we develop the annotation scheme for stigma based on social science theories, and perform a corpus study on the data from Facebook groups on vaccination. The data from pro-vaccination and anti-vaccination discussion groups are annotated by trained annotators and by MTurk annotators. We analyze the annotations using LIWC (Linguistic Inquiry and Word Count) software and TF-IDF in order to identify differentiating features between stigmatizing vs. non-stigmatizing content. Our corpus study lays a valuable foundation in computational modeling of social stigma, as it can serve as validation/interpretation of the social science theories through the prism of laypeople understanding. The annotated corpora can be subsequently used for automatic stigma identification. Moreover, the annotation scheme can be applied to study stigma across different themes and diseases and can be utilized during public health informational campaigns and health interventions.