Virtual Society

Virtual Society is an aid in order to test software using life-like simulations. In comparison with other industries, digitalization has started much slower within our Dutch government. Citizens and companies expect that information from the government is instantaneously available and easily accessible (Van Leeuwen, 2020). Governments have many reasons to meet these expectations through investing, time, effort and money in digital transformation for the public sector. Principles of automation and especially algorithms among which Artificial Intelligence and simulations are indispensable in order to achieve these digital transformations. Primarily Virtual Society is being developed for designers within software engineering and data science. Virtual Society offers simulations within a chain of automated systems, through providing these systems with life events (Van Leeuwen, 2020). These events are modelled within the simulation and executes these. Examples of these life events are births and deaths, leaving the parental home, starting a family, etc. It is crucial to note that this information is completely fabricated: it is synthetic data generated by the simulation. Several elements within the simulation are based on statistics. For example, the CBS (see also is an important resource for Virtual Society. Other elements which provide the foundation for the simulation are based on complex models, such as machine learning models, which can approximate the behavior of automated systems within the actual system being a part of the simulation. Since Virtual Society generates data through its simulations, data cannot be traced back to real personal information. The upside of using synthetic data is that the privacy of citizens is not affected. On top of this, synthetic data is much more accessible for a broader audience. Simulations are a replication (digital twin) of reality. Simulation is a dynamic and complex process. Through Virtual Society we achieve the aforementioned based on life events. This simulation model determines through the use of rules derived from algorithms how the simulation progresses. During the simulation runs we obtain insights into how the situation changes during the simulation time. Many modelling approaches towards life events can be distinguished in literature. For example, the probability of births and deaths can be modelled. During the life cycle of a human being several life events can take place, among which the birth of children, following which specific situations arise within a family. When these are modelled using real statistics, life events arise which replicate reality. The consecutive occurrence of life events create a complex model with many interactions, without actually modelling this complexity. In this sense, Virtual Society aims at designing simulations able to replicate the development of populations. In Baines et al. (2004) it was already discovered that the results of simulation studies, when human factors are accounted for, show a difference of 35% compared to more traditional research when human factors are not accounted for.  

Within Virtual Society simulations are set up using Discrete Event Simulation (DES).DES models the system as a discrete sequence of events in time. Each event takes place at a certain moment in time and marks a change in the status of the system. The subset of DES for simulations within Virtual Society is based on the modelling of processes. Many automated systems within the government are based on process automation. Between consecutive 1events it is assumed that no changes occur in the system (next-event-time progression). This leads to a substantial reduction in computation time compared to for example agent-based modelling. 

The result of the project was a tool in the form of a Proof of Concept showing the possibilities of using DES for the simulation of life evens. The project is still  ongoing.  

Dr. Raymond Hoogendoorn
Senior Researcher
Raymond Hoogendoorn is a senior researcher in the field of data science. In 2012 he completed his PhD cum laude at Delft University of Technology. He did this at the Transport and Planning department of the Faculty Civil Engineering an Geosciences. NWO sponsored his project which was aimed at mathematically modelling driving behaviour in case of adverse conditions (for example during an evacuation).  Raymond completed a post-doc and was Assistant Professor at Delft University of Technology in the field of Intelligent Transport Systems (autonomous vehicles). Currently he conducts research in the field of applied data science for health, urban living and smart-mobility applications.