By Scott Merrill and Eric Clark, University of Vermont

The Social Ecological Gaming and Simulation (SEGS) Lab is a multidisciplinary research lab designed to examine pressing problems in the interacting domains of natural ecology and human society. We work on problems ranging from water quality and energy, to looking at how Covid-19 related social distancing behavior changes in different environments – from the grocery store to the park. We also are engaged in trying to understand the influence of human behavior on the spread of disease in livestock animals. Here we have joined with others around the country as part of the Animal Disease Biosecurity Coordinated Agricultural Project (ADBCAP).

Researchers in the SEGS Lab use a variety of computational tools to examine human behavior. One of our powerful modeling approaches is called agent-based modeling. Real world systems can be simulated using a collection of self-directed decision-making entities called agents. Each agent can act independently based on a set of programmed rules. Agents are embedded in a modeled system, such as the swine industry, and are allowed to interact in space and time to mimic a real world supply chain network. In the case of the swine industry, animals, farms, workers, and trucks that deliver feed and animals can all be programmed as agents, each with their own set of rules to follow.

We can encode disease transmission parameters, such as the probability of infection, or not allowing the disease to be passed between animals if they are isolated in biosecure shelters. Using these computer simulations, we can quickly run thousands of scenarios, each of which play out differently due the randomness of the disease transmission probability. We then use these results to analyze systemwide disease spread dynamics. We can also modify rules to quantify the impact of the change on the system.

For example, we can simulate making all the trucks get disinfected during their delivery route, or asking the simulated workers to use a Danish entry biosecurity system. When we run the model many thousands of times, and allow for randomness, we can get a sense of what is possible and what is likely. The dynamic patterns that emerge can be captured in output variables and analyzed statistically.

Another tool we use that frequently generates excitement is experimental gaming simulations. Here we design games that help us understand human behavior. In general, gathering data on human behavior and biosecurity decisions is hard because we don’t want to put actual people or animals into dangerous or unhealthy situations. An alternative and safer way to study these behavioral responses is to use a computer to simulate a world or a conflict and then create scenarios that force a response by the participants.

For instance, if it takes longer, and you will lose time/money, do you always wash your equipment when you move between buildings? What if you are on the clock and you have to complete a number of chores before the end of the day? If you want to take a short outside break during the day, do you sneak out the back door or do you go through the entire entry/exit protocol, including showering and changing clothes?

Work by the SEGS lab researchers using games and models on the ADBCAP project has provided some fascinating insights. One of the most important things that we have shown is that standard epidemiological models of disease spread are unable to mimic the spread and impact of disease without incorporating human behavioral responses. We have ample evidence that humans and the decisions that we make are frequently irrational and messy. We don’t all act the same way when provided the same information – sometimes we do the right thing, sometimes we don’t. We need to incorporate that messiness into models to get a realistic disease response.

The strategies that we use to communicate risk can dramatically change our society’s willingness to practice and invest in disease prevention. These strategies can dramatically impact the severity of disease outbreaks. If we effectively communicate and act on disease information, we can quickly suppress outbreaks, but the flip side is also true. Ineffective messaging leads to high average disease impacts but with a lot of variability in that severity, meaning that sometimes the disease can be quickly suppressed and other times reaches epidemic levels.

Overall, SEGS work with the ADBCAP project shows that behavior and decision-making is a critical piece for disease modeling, and if we don’t include a human behavioral component we will likely end up with serious errors in our predictions.