An Artificial University Drives Real-World Countermeasures to COVID-19

During the ongoing pandemic, we’ve learned human behavior is a significant factor in the success of non-pharmaceutical interventions. From mask wearing to social distancing, how people act is vital to controlling the spread of the virus in the absence of significantly immune population.

But we don’t have to put people at risk to predict what will work. Instead, we can use computer models and realistic simulations to assist in better understanding how behavior affects disease spread.

Earlier in 2020, under the framework of the COVID-19 Healthcare Coalition, MITRE and several Coalition members began applying the computational power of artificial intelligence (AI) and agent-based modeling to address means to protect populations from such pandemic threats.

“Harnessing new research in computational social science, we created The Artificial University (TAU), a complex model that simulates real life in a university environment, says Andreas Tolk, one of the TAU creators. This simulated world was developed by a diverse team of members from the Center for Mind and Culture (CMAC), the Virginia Modeling Analysis and Simulation Center, MITRE, Simudyne, and TensorX.

TAU is realistic in the way it reflects the diverse racial and socioeconomic subgroups of a university. Its “socially capable people” (actually AI-generated agents) adjust their behavior in response to various pandemic-related constraints and interventions.

The team’s goal was to put this open source model into the hands of university leaders to help them create the best possible policies to keep students, faculty, staff, and surrounding communities safe.

To enable broad, inclusive application, the team released TAU as open source. To date, more than 50 universities have downloaded the source code, and the team continues to collect feedback that will shape future versions and extensions of TAU.

Linking Behavior to Outcomes

TAU’s individual-focused systems not only address spatial-temporal aspects of the pandemic but also social networks and other daily-life factors within a university ecosystem. The model includes school-specific spaces such as lecture halls, dorms, dining halls, sport facilities, and more.

The model helps decision-makers understand many interrelated factors and then identify the interventions with the greatest impact. These could include curriculum-delivery methods, COVID-19 testing protocols such as testing wastewater on campus, or protocols for sporting events.

TAU also captures a university’s values and objectives in the metrics applied to evaluate the outcome of the simulation runs.

Starting Fall 2020, every school issued its own guidance on how to bring students back to campus, including rules on physical distancing to slow virus spread. How is it going?

“Schools have had mixed results on achieving compliance,” notes Tolk. “But TAU can help administrators understand the dynamic social factors that motivate students, including the need to be with other students.”

With this knowledge, university administrators can continually adapt the rules and messaging—i.e., educating students to increase acceptance and compliance.

From Universities to Other Organizations

What makes TAU different from other models currently available?

“We’ve used diverse knowledge to combine standard epidemiological models with social science methods,” says Tolk. “Our models are unique in accounting for diverse human factors such as resistance to public-health recommendations, possible refusal of a vaccine—when they’re available—and more.”

While sharing the TAU model with universities, the team realized that the model could be adapted to the needs of many other organizations. That led to the creation of The Artificial Organization (TAO).

The TAO model is customizable to the circumstances of different organizations—from a mall to a concert hall to an office building campus.

As organizations try to reduce the impact of COVID-19 on their businesses, they’re asking the same questions as universities: What’s the optimum in-person capacity for our space? What’s the impact of increasing the population density? At what level should community case prevalence prompt a preemptive shutdown? Should we offer testing at our sites?

TAO makes it easier for organizations of all shapes, sizes, and demographics to identify high-risk scenarios and assist in determining appropriate policy actions.

“The sheer volume of data we can work with is very powerful,” says Tolk. “Sophisticated models like TAU and TAO can help decision-makers develop processes and protocols to protect people from harm.”

For information about licensing this technology, please contact MITRE’s Technology Transfer Office.

—By Andreas Tolk and Liv Blackmon