While the pandemic was at its peak and new channels were projecting infection rates and charts, did you ever wonder as an analytics professional, what type of modeling approach was behind the analysis? Chances are, that projection you were seeing was a result of agent based simulation modeling. Agent based modeling is a relatively recent method that originated in the early 2000s. Some of the key drivers behind agent-based modeling coming into the picture were advances in technology and computing power since agent-based models are much more demanding in terms of system capability requirements, the proliferation of computer science techniques like object-oriented modeling, UML, and state charts, and the need to develop models for systems that we were not able to capture with traditional simulation modeling approaches of system dynamics modeling and discrete event modeling. Agent-based modeling primarily leverages the behavior of each individual in the system to model the system. So essentially, this is a method of building a model from the bottom up. You identify all the different objects or agents and start defining their behaviors. Eventually, during that process, you would also start connecting those agents through their interactions and relationships. Finally, you would put them into similar environments within which they have their dynamics. All these come together to create the system's global behavior or the model of the entire system. Often there may be hundreds or thousands and, in some cases, even millions of individual agents that need to be incorporated to develop the model of the entire system. The current pandemic is an excellent example of some scenarios where agent-based modeling approaches can be used. This can be understood with the following flow at a very high level. Let us say that we have a city with a population of 1,000,000 people. The city is structured in many different blocks, and let us assume that each block is essentially  5 miles by 5 miles in area and the population is uniformly spread across each block. The other assumption can then be that every person in that block knows everyone who decides within a mile of that person. The model then starts by randomly selecting see 100 people and infecting them. The logic from that point onwards is that if an infectious person contacts or gets in touch with a suspectable person, the letter gets infected, and we assign some form of probability of infection. Then that person who has become infected will essentially go through a latency phase which could be a week or two before they become infectious. After that, those people can spread the infection to the others they come in touch with. There are many steps after this, but this is an excellent example of the application of agent-based modeling, specifically in the current environment. In part, this simplified methodology we just reviewed is popularly known as a susceptible, exposed infectious recovered model or the SEIR model. Before agent-based modeling approaches, the SEIR model was solved using differential equations. However, agent-based modeling allows us to add many additional aspects to this analysis that were impossible to include with the differential equation approach. As you can see from the various steps of the model discussed above. One of the key strengths of the models is their explained ability. Having a basic understanding of how a disease might spread and the course of the disease, you can model the spread straightforwardly.