Ualbany analyzing pandemic messaging

Professor studying effective ways to relay health info through social media posts

New research from the University at Albany offers insight on how government agencies can educate the public and combat misinformation on social media during a pandemic.

Jeannette Sutton, an associate professor and the new director of graduate studies at the Ualbany College of Emergency Preparedness, Homeland Security and Cybersecurity, co-published new research in PLOS ONE that analyzed 150,000 tweets about COVID-19 from official accounts of about 700 state and local agencies between February and April.

The goal was to determine which messaging tactics led to the greatest engagement, measured by the number of retweets. Understanding the messaging features that led to shares could help guide future COVID-19 communication strategies, according to Sutton.

“We’ve been doing research on Twitter messaging for eight years or so,” Sutton said. “Many health agencies, especially at the local level, do not have a high reach to begin with. So, giving them insight into what tactics help draw attention to their posts is extremely helpful.”

To identify effective messaging tactics, Sutton, along with a team of collaborators, first manually sifted through a random sample of 9,000 tweets, looking closely at both message content and other features used in tweets. From there, they used automated methods to code and analyze the remaining tweets.

The analysis revealed that tweets characterizing the impacts of the virus, its spread, and actions that people can take to protect themselves, were all features that most strongly inf luenced message retransmission; no single topic was most successful.

Tweets that included videos, and to a smaller extent, images and hashtags, were more likely to be retweeted.

The date the message was sent made a difference too. For example, messages

published after the national emergency declaration on March 13 were passed on 44 percent more frequently than those sent before the declaration. Tactics such as using exclamation points or presenting content in the form of a question did not appear to promote retweets — a notable difference from other crisis situations in the past.

The researchers also identified a list of Covid-19-related keywords and phrases that were commonly used during the three-month period.

“We have identified consistent themes across different hazards that are effective,” Sutton said. “At the same time, some of the usual common messaging tactics — like using exclamation points or asking a question — were found to be counterproductive

in the first few months of the COVID-19 crisis. This makes sense given that the context of a pandemic differs from a fastmoving threat, like a wildfire or f lood.”

Sutton’s co-investigator is Carter T. Butts at the University of California Irvine. The two are now building on their first study to examine tweets from May to August. They caution that the impact of messaging tactics could change as the pandemic continues to unfold.

“Our reports are not usually completed this quickly. But, we recognized early on into the pandemic that there was an urgency for this type of analysis right now,” Sutton said. “Our research will help agencies reach their target audience during the later stages of this pandemic, or the next public health emergency.”

The project is funded through an ongoing RAPID grant from the National Science Foundation to examine COVID-19 risk communication on social media.