Mining Social Media Messages for HIV Testing and Prevention Communication

PROJECT SUMMARY The spread of the SARS-CoV-2 in the US has created novel problems with methods successfully underway in the parent grant that use cutting edge, Big Data techniques for processing vast and diverse data (text, videos, images) to select actionable and acceptable social media HIV messages that match the HIV biomedical needs of target US counties. These problems include: (a) the messages we are currently selecting are not relevant in the context of the pandemic and the current machine learning methods take many months to optimize; (b) combination SARS-CoV-2/HIV messages would be necessary but the methods to produce them have not been developed; and (c) no methods to multiply scarce or new messages have been validated. This emergency- supplement application requests the resources to offset these problems and to use the opportunity to generate new knowledge about HIV and responses to pandemics. Our new approach will use the same successfully tested techniques to select, generate, and deliver combination messages about HIV (i.e., testing, PrEP, and condom use) and messages about social distancing, testing, prevention, and treatment of the SARS-CoV-2, specifically for Men who have Sex with Men (MSM). The current pandemic also serves as a reminder that vaccination is lacking in MSM; therefore, the supplemental project will integrate messaging about vaccines, including an eventual one against SARS-CoV-2. To counteract the new problems resulting from insufficient numbers of appropriate HIV messages for the SARS-CoV-2 times and to adapt current methods which are too time-consuming and inflexible for messaging in such a changing environment, we need (1) new, rapid methods to identify actionable and acceptable messages; (2) new methods to combine HIV recommendations with emerging public health recommendations related to SARS-CoV-2; and (3) new methods to rapidly multiply the resulting messages for social media. Thus, this funding application will pursue the following: Aim 1. Extend current methods to identify regional needs related to the SARS-CoV-2 pandemic and vaccines, Aim 2. Deploy new rapid methods of message selection for a pandemic, and Aim 3. Experimentally test the effect of the new methods by sending (a) the selected experimental messages (combined with links to local service information) to managers of social media accounts and (b) a random selection of HIV/SARS-CoV-2 messages to a group of control counties. This significant and innovative project will be facilitated by unique team expertise in communication and persuasion, Big Data methods, public health, and Bayesian spatio-temporal modeling, and by the participation of leading institutions in the areas of psychology, public health, and computer science. The project is synergistic with new data collection efforts associated with the American Men Internet Survey, which will now collect SARS-CoV-2 seroprevalence data, which will be supplemented to conduct additional behavioral data for the project proposed in this application.