Skip to content
Main menu hidden.
Published: 2021-04-29

New simulation tool for covid-19 beats epidemiological models

NEWS The simulation tool that is being developed at Umeå University to help decision-makers in their actions to prevent the spread of corona now shows excellent results. “We have tested a large number of scenarios on different aspects of the crisis. Each time, our tool has been more accurate than other models used today", says Frank Dignum, Professor in Socially Aware AI at Umeå University.

Text: Victoria Skeidsvoll

The agent-based social simulation tool simulates the spread of the coronavirus, not just based on statistical data but on how people live, behave and what motivates them. The effectiveness of measures depends for a large part on how people’s behavior changes. The simulation uses a separate decision model for each individual, based on social-psychological theories.

State of the art

These models are combined with a state of the art epidemiological model to create a realistic image of the society during the crisis. “After one year experiencing the worst pandemic ever, we now see that the models predict the best measures and which ones to avoid, every time,” says Professor Frank Dignum. 

The new agent-based social simulation tool shows that measures like quarantine, self-isolation and lockdowns are not always effective. Neither are apps for tracking the virus nor school closures. “There is no universal solution. Every measure must be checked against the state of the pandemic in a country, its infrastructure, and its culture”, says professor Frank Dignum. 

His research group on socially aware AI is working in collaboration with researchers around the world to develop the ASSOCC tool. 
“The underlying models have been constantly refined and more aspects have been incorporated. The epidemiological model is now equal to a state-of-the-art model from epidemiologists from Oxford”, Frank Dignum says. 

With the explainable results of the simulations, the research group can be of great help to decision-makers and authorities. “Our framework will enable introducing effective measures that leave more room for interaction while society, trade, and tourism can conduct their business in a safe environment."

Social behavior

Traditional methods of preventing the development of pandemics based on epidemiological models only have proven not to be adequate.

Such models cannot integrate human behavior, the impact we have on each other, and how this changes behavior over time. This can only be successfully studied in a social simulation environment.” 

The next step is to develop large-scale simulation models. "This requires that we reimplement the models, making them even more efficient and possible to run in high-performance computers.”

Developing systems and technology that protect people’s integrity is also of importance. “The biggest advantage of ASSOCC is that we will be able to explain in detail how to effectively and safely create the best results for human health,” says Professor Frank Dignum.

About the research

Frank Dignum, Professor in Socially Aware AI at Umeå University and initiator of ASSOC. Read more about ASSOC here. Professor Dignum’s new book “Social Simulation for a Crisis”, will be available on Springer. More articles from Umeå university. Swedish broadcast report by journalist Emil Hellerud, TV4 can be found here.