Using AI and Social Listening to Address COVID-19 Misinformation in Mecklenburg County, NC

Background: VIRA, the Vaccine Information Resource Assistant, is a chatbot created in collaboration with Johns Hopkins University and IBM Research. VIRA uses natural language processing and artificial intelligence to answer questions about COVID-19 vaccines with pre-populated responses, which are vetted by a Hopkins vaccine expert. VIRA, launched in June 2021, provides tailored, credible, and empathetic vaccine information using a database of over 220 questions and more than 600 responses in both English and Spanish. By evaluating anonymous conversations, the Hopkins team assesses topics of misinformation and provides social listening insights. The chatbot provides an opportunity to better understand vaccine hesitancy and the impact of misinformation on community health. VIRA draws from the growing body of literature dedicated to the potential of social media to influence vaccine acceptance/hesitancy.

Program background: Partnering with NACCHO, the Hopkins team launched VIRA Local, an opportunity to provide custom chatbots to US health departments at no cost. These chatbots incorporate each health department’s respective branding that reflect their unique communities. Since November 2022, custom chatbots have launched in California, Massachusetts, and North Carolina. Health departments are responsible for communicating about the chatbot within their jurisdictions. VIRA Local also enables social listening analysis of each health department’s chatbot, which provides insights on community-level usage, misinformation, and topics of interest.

Evaluation Methods and Results: We at Mecklenburg County are collecting data about COVID-19 vaccine misinformation through regular social listening reports from the Hopkins team and also by reviewing the comments to all social media posts from the County about COVID-19 from 2022 onward to identify, categorize, and compile data about COVID-19 misinformation. We are analyzing those data, in combination with data reported by Project VCTR and CDC, to craft tailored social media posts and replies to comments to those posts and create timely COVID-19 education and outreach materials. Through continual evaluation of the customized chatbot, we have identified opportunities for improving the chatbot’s responses and the need for increasing efforts of marketing the chatbot. Through continual evaluation of the social media research, we have identified the need for more efficient methods to review the comments and the opportunity for delivering more robust data sets. This evaluation has led to improvements in our review processes and additional marketing channels.

Conclusions: One of the 10 Essential Public Health Services is to “communicate effectively to inform and educate people about health, factors that influence it, and how to improve it.” The best use of social listening data is their application to communication that informs the public to increase health knowledge and encourage self-efficacy in health care.

Implications for research and/or practice: We are already using the social listening data we've collected to inform Mecklenburg County’s COVID-19 communications. This type of research, and the methods to apply this research to practice, can be used by local and state health departments nationwide. Moreover, this collaboration pioneers the use of AI-powered chatbots at local levels, enabling future opportunities to improve access to information across a wide range of public health topics. The Hopkins team is developing a dashboard to share information gathered from VIRA with the public.