Agent Based Models

What is an agent based model?

Agent based modelling (ABM) is simply a system, with autonomous agents (individuals) that make decisions based on a set of rules, that is simulated many times computationally. Because you can determine the parameters within which the agents make choices, you can make the agents act in human-like ways e.g. exhibiting fear, error, or bias. In fact, even fear itself can be contagious. ABM allows you to assess the impact of the agents on the system as a whole. The benefit of using a computational model is that you can run simulations thousands of times to paint a picture that's comparable to epidemic data.

This is why ABM is well suited to demonstrate "complex social networks and the direct contacts between individuals, who adapt their behaviours — perhaps irrationally — based on disease prevalence," according to Joshua M. Epstein. Moreover, ABM is a great tool for communication because it's "rule-based, user-friendly and highly visual."

Source: Nature - Modelling to contain pandemics

Using ABM for COVID-19

ABM are able to express scenarios that people would act differently during a pandemic than they would normally. This allows the model to account for the disease as well as the possibility of mass behavioral changes on many fronts.

Fear actually changes human behavior and therefore changes the way viruses spread. The disease, COVID-19, caused by the novel coronavirus is not the only contagion... Knowing this, Epstien asks leaders to consider four (4!) separate epidemics happening right now?

1. COVID-19 2. FEAR OF THE DISEASE

3. FEAR ABOUT THE ECONOMY 4. FEAR ABOUT THE VACCINE

As there is no vaccine availible, there is no way to eridicate the disease. Thus, behavioural changes such as social distancing are the only way to hold off the disease. When social disancing measures take effect and make the statistics appear less scary, it is easy to fall into a false lull of security. However, without a vaccine, if social distancing measures were relaxed, it would cause a second wave and re-kindle the severity of the pandemic.


To the right are various figures from an Epstein et al. paper titled Coupled Contagion Dynamics of Fear and Disease: Mathematical and Computational Explorations, In this paper they showed that even somewhat low levels of fear-motivated flight could immensely impact contagion dynamics.

Basic Rules of the Model

Agent can only be one of the following at a time: 1) Susceptible to pathogen and fear2) Infected with fear only3) Infected with pathogen only4) Infected with pathogen and fear5)Removed from circulation due to fear and infected with pathogen 6) Removed from circulation due to fear and infected with pathogen 7) Recovered from pathogen and immune to fear
Rate at which individuals self-isolate due to fear and recover from fear and return to circulation:1) Rate of removal to self-isolation of those infected with fear only2) Rate of removal from infection with pathogen3) Rate of removal to self-isolation of those infected with fear and pathogen 4) Rate of recovery from fear and return to circulation
Sources: Brookings - Examining Transmission of Fear and Disease & Politico - Are We Already Missing the Next Epidemic?

Contributors: Collette Patel