NSF RAPID Grant Examines Social Networks' Ability to Slow COVID-19 Spread

9/30/2020 Laura Schmitt, Illinois CS

CS professor Hari Sundaram is collaborating with psychology department researchers on a project that investigates whether discussing either general (curbing COVID-19 disease) or specific (wearing a mask) issues can have important consequences on the spread of risky attitudes and behaviors through a network.

Written by Laura Schmitt, Illinois CS

Illinois CS professor Hari Sundaram is collaborating with psychology professor Dolores Albarracin on an NSF-funded project that leverages digital social networks to promote and spread healthy behaviors in the ongoing battle to curtail the coronavirus pandemic.

The ultimate goal of their work is to generate recommendations and predictive algorithms that public health officials can use to fine-tune their messaging in order to elicit more widespread positive behaviors.  

The project tests the theory that discussing either general (curbing COVID-19 disease) or specific (wearing a mask) issues can have important consequences on the spread of risky attitudes and behaviors through a network.

Both faculty members have a longstanding interest in tackling such collective-action problems that involve complex, large-scale decision-making processes. Sundaram brings his ability to develop computational network models, while Albarracin brings her expertise in measuring attitudes and behavior and the network agenda-setting theory being tested in the grant.

Psychology professor Dolores Albarracin and CS professor Hari Sundaram received an NSF RAPID grant to leverage digital social networks to promote and spread healthy behaviors that combat COVID-19.[cr][lf]<p style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: Calibri, sans-serif;"> </p>[cr][lf]
Psychology professor Dolores Albarracin and CS professor Hari Sundaram received an NSF RAPID grant to leverage digital social networks to promote and spread healthy behaviors that combat COVID-19.

Together, they are analyzing millions of Twitter, Facebook, and Instagram posts to build a statistical model that quantifies and predicts the probability of individuals to mimic a behavior they see on social media.

Computer scientists can build predictive models based on behavior observed from people’s social media posts, said Sundaram. However, these models lack any information about what people were thinking at the time they performed those behaviors.

“What is missing here is information on their attitude,” Sundaram said. “We’ll be able to glean that from the careful experiments that Dolores will do.”

Albarracin and Sundaram will create an artificial social network comprised of nearly 2,000 volunteer participants nationwide who will post and interact on Facebook and Twitter.

“We can manipulate the mix of healthy and risky behaviors in the network and study the impact of discussing either general or specific issues,” Sundaram said. “A general agenda is discussing support of COVID-19 prevention efforts. A specific agenda is discussing support of wearing a mask when in public.”

According to Albarracin, a general agenda could be problematic when people discuss specific behaviors that include risky actions, such as socializing in person without a mask. The general agenda in this case could lead to those negative behaviors being averaged into a riskier result, such as less support for COVID-19 prevention, in general.

“The reason this is problematic is that less support for COVID-19 prevention will then be expressed through a multitude of behaviors like vaccine hesitancy and opposition to business closures when necessary,” she said. “In contrast, a specific agenda such as whether we support wearing masks would encapsulate the negative behavior. Thus, the discussion might lead to less support for mask wearing but would not affect attitudes toward a COVID-19 vaccine.

“However, if all behaviors in the network are healthy, the generalized message would be positive,” Albarracin said.

The project is part of NSF's Rapid Response Research (RAPID) program, which provides one-year grants (maximum $200,000) to conduct non-medical research that can be used immediately to explore how to model and understand the spread of COVID-19, to inform and educate about the science of virus transmission and prevention, and to encourage the development of processes and actions to address this global challenge.

Albarracin is the principal investigator of the project and Sundaram is the co-PI. They are working with computer science graduate student Himel Dev and research assistant professor Sally Chan and post-doctoral researcher Bita Fayaz-Farkhad, who are co-investigators from the department of psychology.


Share this story

This story was published September 30, 2020.