Proceedings of the 2020 Winter Simulation Conference K.-H. Bae, B. Feng, S. Kim, S. Lazarova-Molnar, Z. Zheng, T. Roeder, and R. Thiesing, eds.

Laura H. Kahn

Ian McCulloh

Accenture Federal Services
800 North Glebe Road
Arlington, VA, 22203 USA


U.S. federal government agencies oversee a wide array of citizen benefits, which affect millions of Americans.Federal benefits administration is a complex interaction of systems that can be approximated with a modularframework and digital twin. Rather than focusing on individual elements of the benefits administrationoperation, we aim to minimize interface issues between the elements by modeling the entire operationusing a holistic modular framework. We also present a digital twin discrete event simulation of thebenefits administration system to measure how much new Artificial Intelligence (AI) technologies improvegovernment services.


Federal agencies responsible for adjudicating health, food, financial and other benefits are increasingly askedto deliver services faster and cheaper with fewer resources. Inefficiencies within benefits administration cannegatively impact vulnerable citizens in need of time-sensitive critical services. AI solutions are increasinglybeing used to meet this demand and have consistently been touted as bringing massive amounts of innovationto government. Modeling the benefits administration system as a collection of independent agents thatexecute behaviors for the system they represent allows agencies to customize and explore the dynamicinteractions in the systems they manage with minimal business risk (Bonabeau 2002; Scala et al. 2019).Discrete event simulation has been used to reduce wait times and improve customer service in physicalhealth care and other citizen service settings (Duguay and Chetouane 2007). A modular macroscopic-levelframework with a benefits application system, application intake, decision-maker network and decision-maker processes was created to approximate benefits administration operations. In addition to developingthe macroscopic framework, we quantified the impact of introducing an AI solution into the benefitsapplication system with a digital twin.


A modular macroscopic framework for modeling the federal benefits administration system is presented.Our framework and digital twin can be used to model physical and / or virtual systems and has been adaptedto approximate the interaction and impact of people, processes and new AI technology have in the benefitsadministration context. The framework comprises: the benefits administration system, application intake,a decision-maker network comprising any number of people that make benefits eligibility determinationfor incoming applications, and any number of steps in the decision-making process as shown in Figure 1.Customized variables could be presented in any component of the framework, depending on the specificfederal agency’s context. Benefits administration operations could be simulated and variability approximatedwith discrete event simulation using a general-purpose simulation software like Simio or other types of tools
Macroscopic Model Framework in a Federal Benefits Context. (Dennis and Sturrock 2011). The digital twin includes a Benefits Application System that contains simpleand complex applications sources arriving into the system and routed to either NoAI or AI sources, whichcould include any type of new AI technology, a resource pool of Decision-Makers in a Decision-Makernetwork and a set of agency-defined processes that a Decision-Maker follows to arrive at a decision. Thedigital twin is adaptable to system changes as they occur and could be used as a general approach forimproving efficiencies, operations, and customer service.


A baseline, ‘current state’ digital twin representing the benefits application system was simulated andpermutations were performed to study their effects. Simple and complex benefits applications are theentities that arrive in the system independently of each other at one moment in time and indicate a changeto the system. The simple and complex application entities move from to either a NoAI or AI sourceobject, to any number of Decision Makers servers over a network of connectors and nodes and exits thesystem at the Decision sink, where a decision is made. Throughput and processing time efficiencies weremeasured and gained by the introduction of AI technologies.


Simulation methods serve as proxies to exploring and understanding ‘what if’ scenarios to benefits admin-istration operations using a digital twin. By perturbing the digital twin in the microscopic model, federalagencies can understand how much changes in any part of the system affects the entire system, thereforeimproving the benefits administration system with minimal risk. The modular framework with a federalbenefits context.


Bonabeau, E. 2002. “Agent-based modeling: Methods and techniques for simulating human systems”. Proceedings of the National Academy of Sciences 99(suppl 3):7280–7287.
Dennis, P. C., and D. T. Sturrock. 2011. “Introduction to Simio”. In Proceedings of the 2011 Winter Simulation, edited by K. Bae, S. K. B. Feng, S. Lazarova-Molnar, Z. Zheng, T. Roeder, and R. Thiesing, Volume 1, 29–38. Institute of Electrical and Electronics Engineers, Inc.
Duguay, C., and F. Chetouane. 2007. “Modeling and Improving Emergency Department Systems using Discrete Event Simulation”. Simulation 83(4):311–320.
Scala, P., M. Mujica, D. Delahay, and J. Ma. 2019. “A Generic Framework for Modeling Airport Operations at a Macroscopic Scale”. In Proceedings of the 2019 Winter Simulation Conference, edited by K. Bae, S. K. B. Feng, S. Lazarova-Molnar, Z. Zheng, T. Roeder, and R. Thiesing, Volume 1, 512–523. Institute of Electrical and Electronics Engineers, Inc.