RAPID: Social un-distancing: Understanding self-privacy violations in online communities during the Coronavirus pandemic

The Covid-19 global crisis is unprecedented in a number of ways, one being the scale and scope of human interaction through social media, as people across the world have resorted to online mediums to connect with others. Early evidence indicates that this expanded breadth and depth of online activity may magnify privacy risks for individual users, offering increased opportunity for privacy violations. However, aside from some notable exceptions such as contact tracing apps, online connectedness has not been studied through the lens of privacy risk. This project will investigate how increased disclosure of personal information during the Coronavirus crisis poses unique risks to users’ wellbeing, leaving them vulnerable to privacy violations and subsequent harms that can further worsen the current global health crisis. Investigators will develop and distribute anonymized, annotated COVID-19 related datasets collected from online social platforms in the USA and Italy for the purposes of understanding unique risks to individual privacy posed by COVID-19 crisis. Framing self-disclosure as a strategic and inherently social behavior, investigators will study observed individual and collective rewards for sharing and explore how individual cost/benefit calculations are mediated during the Coronavirus crisis. Outcomes of this project will provide insights into the unique evolution of privacy attitudes during crisis, specifically, how oversharing of personal information is expedited or even encouraged, leaving users vulnerable to privacy breaches and exploits. Project outcomes will provide novel computational methods to identify utterances of self-privacy violations and, critically, to contextualize this risky behavior. These insights will be critical for effectively managing the health and well-being of individuals and communities during COVID-19 and future pandemics. The project will develop convolutional neural networks for labeling of emotional and informational textual utterances of self-disclosure on conversational datasets related to Covid-19 crisis. Semantic labeling approaches, to better capture the language of personal information sharing will also be included in the analysis, for a better modeling effort. These methods will be used to furnish fine-grained labels of instances of self-disclosure in user-centric conversations collected from online social platform. In parallel, the investigators will develop game-theoretic models of self-disclosure in social context. These strategic models will support formal understanding of privacy risk vs. social reward at the individual and collective scales. Data collection, algorithm development and model refinement will move forward in tandem, enabling the most rapid possible response during the Coronavirus crisis. Parallel to a focus on domestic users and English-language text, the investigators will collect and analyze data from Italian social and mainstream media in order to explore the cultural and infrastructural “signatures” of these phenomena, as well as to understand self-disclosure at differing points in the epidemic lifecycle. The dataset with annotations as well as open source code related to the models will be shared with the research community. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.