Identifying Medical Self-Disclosure in Online Communities

Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 4398–4408 June 6–11, 2021. ©2021 Association for Computational Linguistics 4398 Identifying Medical Self-Disclosure in Online Communities Mina Valizadeh∗and Pardis Ranjbar-Noiey∗and Cornelia Caragea and Natalie Parde Department of Computer Science University of Illinois at Chicago {mvaliz2, pranjb3, cornelia, parde}@uic.edu Abstract Self-disclosure in online health conversations may offer a host of benefits, including ear- lier detection and treatment of medical issues that may have otherwise gone unaddressed. However, research analyzing medical self- disclosure in online communities is limited. We address this shortcoming by introducing a new dataset of health-related posts collected from online social platforms, categorized into three groups (NO SELF-DISCLOSURE, POSSI- BLE SELF-DISCLOSURE, and CLEAR SELF- DISCLOSURE) with high inter-annotator agree- ment (κ = 0.88). We make this data available to the research community. We also release a predictive model trained on this dataset that achieves an accuracy of 81.02%, establishing a strong performance benchmark for this task. 1 Introduction Self-disclosure is a communicative act that helps people develop close relationships (Altman and Taylor, 1973) through reciprocal sharing of per- sonal information, promoting maintenance of trust and security (Bruss and Hill, 2010). It is defined as the “process of making the self known to others” (Joinson and Paine, 2007), often by sharing one’s personal thoughts, opinions, or experiences. For example: • When I was 19 years old, I met a man on the internet. He was 21 years old, 2 years older than me. • My name is Amy and I live in Australia. • I have suffered from migraines for three years. In addition to facilitating social bonds, self- disclosure in general produces a wide variety of health benefits and plays a critical role in success- ful treatment of many physical and psychological ∗Authors contributed equally. health issues (Ellis and Cromby, 2012). The reve- lation of private and sensitive information is more widespread online than in face-to-face interactions (Joinson, 2001; Tidwell and Walther, 2002; Wang et al., 2016), perhaps due to the anonymity that online platforms provide, or the ability to avoid the face-to-face stigma of some uncomfortable top- ics. The benefits of medical self-disclosure (i.e., disclosing symptoms, diagnoses, or other informa- tion specifically related to mental or physical health issues) in online settings may be particularly valu- able from a clinical perspective, enabling earlier detection and treatment of medical issues that may have otherwise gone unaddressed (Pennebaker and Chung, 2007; Joinson, 2001). However, medical self-disclosure has been under-explored in prior computational work. We set out to address that limitation, making several key contributions. First, we establish the novel task of medical self- disclosure detection, and create a 6,639-instance dataset comprised of public online social posts covering a wide range of mental and physical health issues, annotated with graded (NO SELF- DISCLOSURE, POSSIBLE SELF-DISCLOSURE, and CLEAR SELF-DISCLOSURE) labels. We re- lease this dataset to the research community to facil- itate easy replication of our work, as well as rapid entry …

Content truncated to 500 words