HIV risk on Twitter: The ethical dimension of social media evidence-based prevention for vulnerable populations

Abstract

As of 2016 the HIV/AIDS epidemics is still a key public health problem. Recent reports showed that alarmingly high numbers of people in vulnerable populations are not reached by preventative efforts. Despite technology improvement, we are not yet able to identify populations that are most susceptible to HIV infections. In order to enable evidence-based prevention, we are studying new methods to identify HIV at-risk populations, exploiting Twitter posts as possible indicators of HIV risk. Our research on social network analysis and machine learning outlined the feasibility of using tweets as monitoring tool for HIV-related risk at the demographic, geographical, and social network level. However, this approach highlights ethical dilemmas in three different areas: data collection and analysis, risk inference through imperfect probabilistic approaches, and data-driven prevention. We contribute a description, analysis and discussion of ethics based on our 2-year experience with clinicians, IRBs, and local HIV communities in San Diego, California.