© 2024. All rights reserved.
© 2024. All rights reserved.
In today’s world, human behavior, social networks, and civil infrastructures are closely intertwined. Coupled social, technical, informational, and organizational networks (STIOs) do not stand-alone. They (i) consist of many interacting physical, technological, and human or societal components, (ii) are spatially distributed, managed by different federal, state, or commercial entities, and (iii) operate at multiple time scales. Coupled STIO networks can have multiple layers of networks, each driven by a potential distinct social, physical, biological or economic theory and capturing the interaction at a chosen scale. Examples of such systems include regional transportation systems, regional electric power markets and grids, the Internet, ad-hoc telecommunication, communication and computing systems, content delivery networks, social networks, search networks and public health services. What all of these systems have in common is that they are networked – individual agents or components interact only with a specified set of components. The links in such networks can be physically real or a matter of convention, such as those imposed by law or social norms, depending on the specific system being represented. Thus, STIO networks consist of one or more social networks interacting with underlying technological and physical networks.
Reasoning about large STIO networks is challenging and novel, since unlike physical systems, STIO networks are affected not only by physical laws but also by human behavior, regulatory agencies, courts, government agencies, and private enterprises. The urban transportation system is a canonical example of such interaction: traffic rules in distant parts of the city can have an important bearing on the traffic congestion in downtown, and seemingly “reasonable” strategies such as adding a new road somewhere might in fact worsen the congestion. The complicated inter-dependencies within and among various socio-technical systems, and the need to develop new tools, are highlighted by the failure of the electric grid in the Northeastern US in 2003. The massive power outage left people in the dark along a 3,700 mile stretch through portions of Michigan, Ohio, Pennsylvania, New Jersey, New York, Connecticut, Vermont, and Canada. Failure of the electric grid led to cascading effects that slowed down Internet traffic, closed financial institutions, and disrupted transportation; the New York subway system came to a halt, stranding more than 400,000 passengers in tunnels.
Global-scale events that have received attention in the media as well as academic circles are current and past financial crises, collapse of the coupled infrastructure system caused by the power grid failure, e.g., the Northeast blackout of 2003, pandemics caused by influenza-like illnesses and SARS-CoV-2, national security problems posed by stateless adversaries and terrorist organizations, and global security and sustainability problems arising from potential climate change and population growth. Individuals, institutions, and governments could not prevent the Northeast blackout. We are struggling as a society to combat and control global terrorism. Finally, reducing the economic burden and human suffering of global pandemics such as COVID-19 continues to be a challenge although a certain amount of global coordination and cooperation was on display. And in the case of climate change, a clear consensus is yet to emerge as to the course of action and potential impacts.
Informally, we use the term digital twin to refer to a set of virtual constructs that mimic the structure, context, and behavior of coupled STIO networks. We will also use the term synthetic information (SI), digital twin or digital similars interchangeably in our discussion – we will discuss the use of these terms below. Our focus here is representing and modeling to support evidence-based decision making, situation assessment, and forecasting as it pertains to real-world coupled STIO networks. The term synthetic data appears to be first used by Rubin in the context of US Census data. The origin of the term digital twin is attributed to David Gelernter in his 1991 book Mirror Worlds. We have used synthetic data/information as well as digital twins in our prior work. The digital twin can be continually updated by data corresponding to its real-world counterpart. Researchers in the engineering community have used “fit for purpose multi-scale models” to denote a similar concept. The concept of synthetic information and digital twins are similar but in our opinion have important differences.
There is another sense in which the term synthetic data is used that is perhaps a bit different than a digital twin. This has to do with separation of concerns. Digital twins, as defined in a recent National Academies report (December 2023), consist of the data structure and the associated models that might use the twin. Synthetic information, in contrast, focuses on the underlying data structure. The algorithms and models that use the synthetic networks as a fabric for various spreading processes are not usually included in the definition of synthetic information. Researchers have also used “live digital twins” or “live synthetic information” to denote the fact that the virtual system is evolving in sync with the real system that it is intending to represent. Finally we propose another term, namely counterfactual dynamic digital twin, that refers to dynamic data that can be viewed as an extrapolation of the data in time and space based on certain counterfactuals (scenarios). Of course, this brings other issues to the fore, including what aspect of the system is evolving synchronously with the underlying system, what sort of time delays exist between the two, and how one accounts for these differences.
The term synthetic here has two distinct meanings; that the data is synthesized by integrating diverse data sources combined with appropriate statistical and machine learning methods for (i) disaggregation, (ii) extrapolation, and (iii) interpolation. The other use of the term synthetic is to convey the idea that the constituent elements in the data do not match any real human agent. Matching with infrastructure elements is desirable and is often possible depending on the availability of the data. In a nutshell, socio-technical digital twins provide a socially and geographically explicit coordinate system that can be used for inter- and intra-agent interactions.
Our group has been developing methods for creating digital twins of coupled STIO networks for over 30 years. The work started as part of the TRANSIMS [BB+1999] project at the Los Alamos National Laboratory (LANL). Richard Beckman and his colleagues [BBM1996] proposed a methodology for constructing synthetic populations and this methodology has become a classic. Over the years, we added several new ideas, algorithms. and extensions as part of a number of projects, including the NISAC-UIS project at LANL. We continued the work at Virginia Tech, funded by a number of NSF, NIH, and DTRA projects, particularly NIH MIDAS and DTRA CNIMS. See our tutorials [MS2016, MS2017] for details on this subject. See [BES2005, LA+2024a, LA+2024b, MM+2014, MV2013, RL+2015] for overview articles on synthetic information, application to policy, and mathematical foundations of digital twins and agent-based models that we have written over the past 30 years.
Our team’s work was also highlighted in two articles by science writer Mitch Waldrop. These focused on the role of agent-based models in advising policy makers related to supporting the DoD’s Ebola response efforts and large-scale human-initiated crises.
To support the federal government in their response to the COVID-19 pandemic, we created digital twins of social contact networks of the state of Virginia, as well as other parts of the United States (US). Some examples of our other work include:
Let us discuss the digital twins of US populations and associated STIO networks in some detail.
A synthetic agent (e.g., person) is assigned states and interactions that make it statistically consistent with members of the (real) population without necessarily matching the characteristics of any specific (real) person. A synthetic population represents a set of synthetic agents (e.g., people) that share common geographic, social or biological characteristics (e.g., people in a rural or urban region, individuals from a given tribe). These populations and networks are formed by collecting a large and diverse set of publicly and/or commercially available datasets. These datasets include census, land use, mobility, activity, behavioral and transportation surveys, and building maps. The datasets have been integrated in a first principles manner to construct these synthetic populations. They have been used for highly accurate, national level, agent-based modeling tasks.
The synthetic data is formed by taking the empirical distribution of particular attributes within an area (for example, the number of people per household, age, or income), then iteratively forming a population that matches those distributions while also preserving observed associations between those attributes. We apply a similar process to mobility data. This process means that there is no direct correspondence between any real person and any single synthetic agent within our synthetic population.
A synthetic social contact network is formed by viewing each agent as a vertex and adding an edge between two vertices if they are in physical proximity to each other. The notion of proximity is often dependent on the application. Edges and nodes can have additional attributes, and in this sense these networks can be viewed as knowledge graphs as well; see [BB+2009, HF+2017, MS2016, MS2017]. Agents can be humans, animals, mobile devices, or cells in a body.
As an extension of the first example, we assign each individual infectious disease information. The social contact network edges are endowed with information related to disease transmission. Agent-based simulations are then used to create a synthetic disease outbreak. This creates a new layer, called synthetic pandemic or synthetic epidemic. Creation of the new layer needs: (i) models of disease specific properties associated with each individual; (ii) initialization conditions on how the outbreak began; and (iii) pharmaceutical and non-pharmaceutical interventions that might be in place. See [EG+2004] for how this was used in the context of an important study related to a bio-terrorism event. A particularly important extension we incorporated during our COVID-19 response efforts was to create live digital twins – this required the ingestion of near real-time data pertaining to disease dynamics [BC+2023]. In other words, we aimed to bring the digital twin to life, contextualizing it with current ground conditions.
Basic synthetic information can be extended with new data sources. To illustrate this, see [TB+2023] where we discuss how we can create synthetic data that captures household-level electrical energy demand profiles. Using a number of new data sources, including energy consumption, housing structure types and building materials, weather, and irradiance combined with models of individual behavior within a house (note that in the previous example, the activities of individuals within a household were not considered), we create hourly demand for energy over an entire normative year for every household in the US.
Another extension is a synthetic network that is a digital twin of an electric power network. The comprehensive dataset consists of nodes with attributes such as geocoordinates; type of node (residence, transformer, or substation); and edges with attributes, such as geometry, type of line (feeder lines, primary or secondary), and line parameters as well as household-level energy demand described above. See [TB+2023] for additional details on creating a digital network of the distribution grid of a region in the US.
As a more complicated example and an extension of the above digital networks, we briefly discuss a multi-layered, multi-theory, multi-fidelity STIO network that spans an urban region. The network consists of several layers, each representing constituent elements of the urban infrastructure. There are multiple theories at play that form the basis of constructing the network, which is often described at different levels of detail. Individual human agents in the networks and associated human networks that interact with that physical network are also represented in this process, creating a complex layered networked system. A first version of the network was created as part of the NISAC UIS project at Los Alamos. This complex STIO network thus represents a functioning urban-scale infrastructure system. See [PH+2016, SL+2013] for further details.
Our methodology ensures that privacy is maintained. We apply our synthesizing process to data that is already public, and in the case of US Census data, has already had privacy enhancing methods applied to it. In the case of the 2020 US Census, the data that was released to the public had differential privacy applied to it. Because differentially private data is known to be immune to post-processing, this means that we have a mathematical guarantee that this aspect of our synthetic data is private. Furthermore, when we use sources that have not been subjected to differential privacy processing, we rely on data sources that are already public.
In addition to the privacy and availability benefits, synthetic populations permit a larger degree of control over conditions. Since the populations are synthesized, one can create many variations of the population and network, and in doing so, design a controlled experiment where the “ground truth” is completely understood. Indeed, synthetic population data is well-suited to this form of challenge. In 2015, similar techniques were used as part of an Ebola forecasting competition; an agent-based disease model for Ebola running on a synthetic population was used to generate partial disease outcomes.
The COVID-19 pandemic has emphasized the need for a more robust disease surveillance infrastructure. However, the development of such an infrastructure runs into a double-bind. Assurances of privacy require adversarial testing against realistic systems. Likewise, it is difficult to build realistic systems without access to data. To get around this problem, we created synthetic datasets that reflect a realistic disease outbreak. This data was used as a component in the US-UK Prize Challenge on Privacy-Enhancing Technologies. In the challenge, participants were tasked with creating personalized risk forecasts of infection in a privacy-preserving manner. The challenge was put on by the UK’s Center for Data Ethics and Innovation (CDEI) and Innovate UK, as well as by the US National Institutes of Standards and Technology (NIST), and the National Science Foundation (NSF) in cooperation with the White House Office of Science and Technology Policy (OSTP). See our NSF PREPARE website for more information.
The synthetic data produced by our methods can be used in several ways. We have demonstrated its value and discussed a few examples above; in a nutshell the data can be viewed as a socially and spatially explicit data structure. Emerging areas of AI application include:
Complex social systems require detailed, data-driven models at scale to enable forecasting, planning, and intervention modeling. Synthetic information systems resolved at an individual level provide a natural data structure to support AI and analytics at scale.
The synthetic information systems we have created are most useful when combined with individual-level dynamic models, aka agent-based models. Agent-based models are very well suited to use such synthetic information and create new classes of synthetic information. In conjunction with the synthetic information described above, we have created agent-based models and simulations to produce dynamical outcomes. Agent-based models can be thought of as a way to compose individual-level algorithmic representations to form composed dynamics. We describe a few salient modeling environments that use synthetic information to support policy makers over the last 30 years:
We describe a few of them here.
Agent-based models provide a way to understand the composition of individual-level dynamics and associated feedback that are relevant to policy making, but they also provide a new way to produce live digital twin data. The live digital twin data can be used in several additional ways, including:
We have done work in each of these areas over the last three decades; see the selected publications below.
We have been actively developing synthetic information products for various communities for 30 years. Additionally, we have been developing high performance computing-oriented simulation and analytical methods during that time. Illustrative publications include:
[BBM1996] Beckman R, Baggerly K, McKay M. Creating synthetic baseline populations. Transportation Research Part A: Policy and Practice. 1996 Nov 1;30(6):415-29. Link
[BB+2009] Barrett C, Beckman R, Khan M, Vullikanti A, Marathe M, Stretz P, Dutta T, Lewis B. Generation and analysis of large synthetic social contact networks. Proceedings of the 2009 Winter Simulation Conference (WSC). 2009 Dec 13;1003-1014. Link
[EV+2004] Eubank S, Vullikanti A, Marathe M, Srinivasan A, Wang N. Structural and algorithmic aspects of massive social networks. Proceedings of the 15th Annual ACM-SIAM Symposium on Discrete Algorithms (ACM SODA). 2004 Jan 11, 718-727. Link
[HC+2023] Harrison G, Chen J, Mortveit H, Hoops S, Porebski P, Xie D, Wilson M, Bhattacharya P, Vullikanti A, Xiong L, Marathe M. Synthetic Data To Support US-UK Prize Challenge For Developing Privacy Enhancing Methods: Predicting Individual Infection Risk During A Pandemic. 2023. Data set. Link
[HF+2017] Hasan S, Fox E, Bisset K, Marathe M. EpiK: A Knowledge Base for Epidemiological Modeling and Analytics of Infectious Diseases. Journal of Healthcare Information Research. 2017 Dec;1(2):260-303. DOI: 10.1007/s41666-017-0010-9. Link
[LC+2016] Lum K, ChungBaek Y, Eubank S, Marathe M. A two stage fitted values approach to activity matching. International Journal of Transportation. 2016;4(1):41-56. Link
[MS+2024] Mehrab Z, Stundal L, Venkatramanan S, Swarup S, Lewis B, Mortveit M, Barrett C, Pandey A, Wells C, Galvani A, Singer B, Leblang D, Colwell R, Marathe M. Network Agency: An Agent-based Model of Forced Migration from Ukraine. To appear in Proceedings of the International Conference on Autonomous Agents and Multiagent Systems (AAMAS 2024).
[MV+2024] Meyur R, Vullikanti A, Swarup S, Mortveit H, Centeno V, Phadke A, Poor HV, Marathe M. Ensembles of realistic power distribution networks. Proceedings of the National Academy of Sciences. 2022 Oct 18;119(42):e2205772119. Link
[SP+2017] Subbiah R, Pal A, Nordberg E, Marathe A, Marathe M. Energy demand model for residential sector: a first principles approach. IEEE Transactions on Sustainable Energy. 2017 Feb 15;8(3):1215-24. Link
[TB+2023] Thorve S, Baek Y, Swarup S, Mortveit H, Marathe A, Vullikanti A, Marathe M. High resolution synthetic residential energy use profiles for the United States. Scientific Data. 2023 Feb 6;10(1):76. Link
[TD+2023] Tozluoğlu Ç, Dhamal S, Yeh S, Sprei F, Liao Y, Marathe M, Barrett C, Dubhashi D. A synthetic population of Sweden: datasets of agents, households, and activity-travel patterns. Data in Brief. 2023 Jun 1;48:109209. Link
[XN+2015] Xia H, Nagaraj K, Chen J, Marathe M. Synthesis of a high resolution social contact network for Delhi with application to pandemic planning. Artificial Intelligence in Medicine. 2015 Oct 1;65(2):113-30. Link
[BB+2010a] Barrett C, Beckman R, Channakeshava K, Huang F, Vullikanti A, Marathe A, Marathe M, Pei G. Cascading failures in multiple infrastructures: From transportation to communication network. 2010 5th International Conference on Critical Infrastructure (CRIS). 2010 Sep 20, 1-8. Link
[BB+1999] Barrett C, Beckman R, Bisset K, Berkbigler K, Bush B, Eubank S, Hurford J, Konjevod G, Kubicek D, Marathe M, Morgeson J, Rickert M, Romero P, Smith L, Speckman M, Speckman P, Stretz P, Thayer G, Williams M. TRANSIMS (TRansportation ANalysis SIMulation System): Volume 0: Overview, LA-UR-99-1658; Volume 2: Software, Part 1: Modules, LA-UR-99-2574; Volume 2: Software, Part 2: Selectors, LA-UR-99-2575; Volume 2: Software, Part 3: Test Networks, LA-UR-99-2576; Volume 2: Software, Part 5: Libraries, LA-UR-99-2578; Volume 3: Files, LA-UR-99-2579; Volume 6: Installation, LA-UR-99-2580. 1999.
[BB+2013] Barrett C, Bisset K, Chandan S, Chen J, Chungbaek Y, Eubank S, Evrenosoğlu Y, Lewis B, Lum K, Marathe A, Marathe M. Planning and response in the aftermath of a large crisis: An agent-based informatics framework. Proceedings of the 2013 Winter Simulations Conference (WSC). 2013 Dec 8, 1515-1526. Link
[BJ+2018] Barrett C, Johnson J, Marathe M. High Performance Synthetic Information Environments An integrating architecture in the age of pervasive data and computing: Big Data (Ubiquity Symposium). Ubiquity. 2018 Mar 8;2018(March):1-1. Link
[BC+2013] Beckman R, Channakeshava K, Huang F, Kim J, Marathe A, Marathe M, Saha S, Pei G, Vullikanti A. Integrated Multi-Network Modeling Environment for Spectrum Management. IEEE Journal on Selected Areas in Communication, special issue on Network Science. 2013;31(6):1158-1168. Link
[BC+2023] Bhattacharya P, Chen J, Hoops S, Machi D, Lewis B, Venkatramanan S, Wilson ML, Klahn B, Adiga A, Hurt B, Outten J. Data-driven scalable pipeline using national agent-based models for real-time pandemic response and decision support. The International Journal of High Performance Computing Applications. 2023 Jan;37(1):4-27. Finalist Gordon Bell Special Prize for HPC-Based COVID-19 Research. Link
[BC+2014a] Bisset K, Chen J, Deodhar S, Feng X, Ma Y, Marathe M. Indemics: An interactive high-performance computing framework for data-intensive epidemic modeling. ACM Transactions on Modeling and Computer Simulation (TOMACS). 2014 Jan 1;24(1):1-32. Link
[CS+2013] Chandan S, Saha S, Barrett C, Eubank S, Marathe A, Marathe M, Swarup S, Vullikanti A. Modeling the interaction between emergency communications and behavior in the aftermath of a disaster. Social Computing, Behavioral-Cultural Modeling and Prediction: 6th International Conference, SBP 2013. Washington, DC, April 2-5, 2013. 476-485. Springer Berlin Heidelberg. Link
[CB+2014] Channakeshava K, Bisset K, Marathe M, Vullikanti A. Reasoning about mobile malware using high performance computing based population scale models. Proceedings of the Winter Simulation Conference. 2014 Dec 7, 3048-3059. Link
[EG+2004] Eubank S, Guclu H, Vullikanti A, Marathe M, Srinivasan A, Toroczkai Z, Wang N. Modelling disease outbreaks in realistic urban social networks. Nature. 2004 May 13;429(6988):180-4. Link
[IC+2023] Islam K, Chen D, Marathe M, Mortveit H, Swarup S, Vullikanti A. Towards Optimal and Scalable Evacuation Planning Using Data-driven Agent Based Models. Proceedings of the International Conference on Autonomous Agents and Multiagent Systems (AAMAS). 2023;2397-2399. Link
[IM+2022] Islam K, Marathe M, Mortveit H, Swarup S, Vullikanti A. Data-driven Agent-based Models for Optimal Evacuation of Large Metropolitan Areas for Improved Disaster Planning. Proceedings of the 21st International Conference on Autonomous Agents and Multiagent Systems. 2022 May. Link
[IM+2023] Islam K, Meyur R, Kishore, A, Thorve, S, Chen, D, Marathe, M. OptICS-EV: A Data-Driven Model for Optimal Installation of Charging Stations for Electric Vehicles. Proceedings of the International Conference on Computational Science (ICCS). 2023;70-85. DOI: 10.1007/978-3-031-36027-5_6. Link
[MV+2019] Meyur R, Vullikanti A, Marathe M, Pal A, Youssef M, Centeno V. Cascading effects of targeted attacks on the power grid. Complex Networks and Their Applications VII: Volume 1 Proceedings The 7th International Conference on Complex Networks and Their Applications. 2019;155-167. Springer International Publishing. Link
[PS+2013] Parikh N, Swarup S, Stretz P, Rivers C, Lewis B, Marathe M, Eubank S, Barrett C, Lum K, Chungbaek Y. Modeling human behavior in the aftermath of a hypothetical improvised nuclear detonation. Proceedings of the 2013 International Conference on Autonomous Agents and Multi-agent Systems. 2013 May 6;949-956. Link
[QE+2022] Qiu Z, Espinoza B, Vasconcelos VV, Chen C, Constantino S, Crabtree S, Yang L, Vullikanti A, Chen J, Weibull J, Basu K. Understanding the coevolution of mask wearing and epidemics: A network perspective. Proceedings of the National Academy of Sciences. 2022 Jun 28;119(26):e2123355119. Link
[BB+2007] Barrett C, Bisset K, Chen J, Eubank S, Lewis B, Vullikanti A, Marathe M, Mortveit H. Effect of public policies and individual behavior on the co-evolution of social networks and infectious disease dynamics. Proceedings of the DIMACS Workshop on Computational Methods for Dynamic Interaction Networks. 2007 Sept. Link
[BC+2014b] Barrett C, Centeno V, Eubank S, Yaman Evrenosoğlu C, Marathe A, Marathe M, Mishra C, Mortveit H, Pal A, Phadke A, Thorp J. Impact of a surface nuclear blast on the transient stability of the power system. Critical Information Infrastructures Security: 9th International Conference, CRITIS 2014. Limassol, Cyprus. October 13-15, 2014. Revised Selected Papers. 2016;153-158. Springer International Publishing. Link
[BC+2012] Barrett C, Channakeshava K, Huang F, Kim J, Marathe A, Marathe M, Pei G, Saha S, Subbiah B, Vullikanti A. Human initiated cascading failures in societal infrastructures. PloS one. 2012 Oct 31;7(10):e45406. Link
[BE+2013] Barrett C, Eubank S, Evrenosoglu CY, Marathe A, Marathe M, Phadke A, Thorp J, Vullikanti A. Effects of hypothetical improvised nuclear detonation on the electrical infrastructure. International ETG-Congress 2013; Symposium 1: Security in Critical Infrastructures Today. 2013 Nov 5, 1-7. Link
[B+2023] Borchering, et al. Impact of SARS-CoV-2 vaccination of children ages 5–11 years on COVID-19 disease burden and resilience to new variants in the United States, November 2021–March 2022: A multi-model study. The Lancet Digital Health. 2023;17:100398. Link
[BR+2023] Brower A, Ramesh B, Islam K, Mortveit H, Hoops S, Vullikanti A, Marathe M, Zaitchik B, Gohlke J, Swarup S. Augmenting the Social Vulnerability Index using an agent-based simulation of Hurricane Harvey. Computers, Environment and Urban Systems. 2023 Oct 1;105:102020. Link
[CV+2021] Chen J, Vullikanti A, Santos J, Venkatramanan S, Hoops S, Mortveit H, Lewis B, You W, Eubank S, Marathe M, Barrett C, Marathe A. Epidemiological and Economic Impact of COVID-19 in the US. Scientific Reports. 2021;11(20451):1-12. doi.org/10.1038/s41598-021-99712-z Link
[HC+2023] Howerton E, Contamin L, Mullany L, Qin M, Reich N, Bents S, … Marathe M, Lessler J. Evaluation of the US COVID-19 Scenario Modeling Hub for informing pandemic response under uncertainty. Nature Communications. 2023;14(1):7260. Link
[PW+2023] Pandey A, Wells CR, Stadnytskyi V, Moghadas S, Marathe M, Sah P, Crystal W, Meyers L, Singer B, Nesterova O, Galvani A. Disease burden among Ukrainians forcibly displaced by the 2022 Russian invasion. Proceedings of the National Academy of Sciences. 2023 Feb 21;120(8):e2215424120. Link
[PH+2016] Parikh N, Hayatnagarkar HG, Beckman RJ, Marathe MV, Swarup S. A comparison of multiple behavior models in a simulation of the aftermath of an improvised nuclear detonation. Autonomous Agents and Multi-agent Systems. 2016 Nov;30:1148-74. Link
[PT+2011] Phadke A, Thorp J, Barrett C, Eubank S, Marathe A, Marathe M, Vullikanti A. Effects of Multiple Local Network Insults: Vulnerabilities, Analysis and Recommendations. NDSSL-TR 11-001. 2011. Link
[SG+2023] Sherratt K, Gruson H, Johnson H, Niehus R, Prasse B, Sandmann F, … Marathe M, Funk S. Predictive performance of multi-model ensemble forecasts of COVID-19 across European nations. eLife. 2023;12, e81916. Link
[SL+2013] Swarup S, Lum K, Barrett CL, Bisset K, Eubank S, Marathe M, Stretz P. A synthetic information approach to urban-scale disaster modeling. Proceedings of the 2013 IEEE 16th International Conference on Computational Science and Engineering. 2013 Dec 3, 1105-1112. Link
[TS+2022] Truelove S, Smith C, Qin M, Mullany L, Borchering R, Lessler J, Shea K, Howerton E, Contamin L, Levander J, Salerno J, et al. Projected resurgence of COVID-19 in the United States in July—December 2021 resulting from the increased transmissibility of the Delta variant and faltering vaccination. eLife. 2022;11:e73584. Link
[BB+2010b] Barrett C, Bisset K, Leidig J, Marathe A, Marathe M. An integrated modeling environment to study the co-evolution of networks, individual behavior and epidemics. AI Magazine. 2010 Jan 3;31(1):75-87. Link
[BE+2015] Barrett C, Eubank S, Marathe A, Marathe M, Swarup S. Synthetic information environments for policy informatics: a distributed cognition perspective. Governance in the Information Era. 2015 Feb 20, 285-302. Routledge. Link
[BES2005] Barrett C, Eubank S, Smith J. If smallpox strikes Portland… Scientific American. 2005 Mar 1;292(3):54-61. Link
[BE+2004] Barrett C, Eubank S, Vullikanti A, Marathe M. Understanding large scale social and infrastructure networks: A simulation-based approach. SIAM News. 2004;37(4):1-5. Link
[CE+2020] Chen J, Eubank S, Levin S, Mortveit H, Venkatramanan S, Vullikanti A, Marathe M. Networked Epidemiology for COVID-19. SIAM News. July 2020. Link
[CH+2022] Chen J, Hoops S, Bhattacharya P, Machi D, Marathe M. Multi-agent Simulations and Vaccine Allocation Strategies. Exploring the Role of Artificial Intelligence, Network Science, and High-Performance Computing. SIAM News. 2022 January. Link
[LA+2024a] Laubenbacher R, Adler F, An G, Castiglione F, Eubank S, Fonseca L, Glazier J, Helikar T, Jett-Tilton M, Kirschner D, Macklin P, Mehrad B, Moore B, Pasour V, Shmulevich I, Smith A, Voigt I, Yankeelov T, Ziemssen T. Forum on immune digital twins: a meeting report. NPJ Systems Biology and Applications. 2024 Feb 16;10(1):19. Link
[LA+2024b] Laubenbacher R, Adler F, An G, Castiglione F, Eubank S, Fonseca L, Glazier J, Helikar T, Jett-Tilton M, Kirschner D, Macklin P, Mehrad B, Moore B, Pasour V, Shmulevich I, Smith A, Voigt I, Yankeelov T, Ziemssen T. Toward Mechanistic Digital Twins: Some Use Cases. Frontiers in Digital Health. 2024 Mar 6;6. Link
[LS+2013] Lewis B, Swarup S, Bisset K, Eubank S, Marathe M, Barrett C. A simulation environment for the dynamic evaluation of disaster preparedness policies and interventions. Journal of public health management and practice: JPHMP. 2013 Sep;19(0 2):S42. Link
[MM+2014] Marathe M, Mortveit H, Parikh N, Swarup S. Prescriptive analytics using synthetic information. Emerging Methods in Predictive Analytics: Risk Management and Decision-Making 2014. 1-19. IGI Global. Link
[MV2013] Marathe M, Vullikanti A. Computational Epidemiology. Communications of the ACM (CACM). 2013 Jul 1;56(7):88-96. Link
[RL+2015] Ramakrishnan N, Lu C, Marathe M, Marathe A, Vullikanti A, Eubank S, Roan M, Brownstein J, Summers K, Getoor L, Srinivasan A. Model-Based Forecasting of Significant Societal Events. IEEE Intelligent Systems. 2015 Sep 1;30(05):86-90. Link
[BE+2009] Bisset K, Eubank S, Feng X, Marathe M. EpiSimdemics: An efficient algorithm for simulating the spread of infectious disease over large realistic social networks. HPC Users Forum, April 2009, Roanoke, VA.
[MRV2016] Marathe M, Ramakrishnan N, Vullikanti A. Computational Epidemiology and Public Health Policy Planning. A 4-hour tutorial that provides an overview of the state-of-the-art in computational epidemiology from a multi-disciplinary perspective. 30th Annual conference on Artificial Intelligence (AAAI). Phoenix, AZ, 13 February 2016.
[MS2017] Marathe M and Swarup S. Generating Synthetic Populations for Social Modeling. A 4-hour tutorial on the foundations and methods for generating synthetic agents. Autonomous Agents and Multi-agent Systems International conference (AAMAS), São Paulo, Brazil, 8-12 May 2017.Link
[MS2016] Marathe M and Swarup S. Generating Synthetic Populations for Social Modeling. A 4-hour tutorial on the foundations and methods for generating synthetic agents. IJCAI, New York NY, 9-15 July 2016. Link