Agent-Based Modeling: simulations for a better understanding of society
By: Frank Pijpers
The complex society
A busy shopping street or a train station or highway around rush hour sometimes seem like utterly uncoordinated chaos. Yet our society does not consist of a disjointed collection of individuals or businesses. The preferences and choices for what to do and what not to do of individuals, or households, or businesses, lead to all sorts of interactions among them, which ultimately manifest themselves in all sorts of social trends and social relationships. Within the fields of sociology and economics, there is a great deal of descriptive and deductive research that tries to make sense of those underlying interactions from data, such as official statistics or targeted surveys or even small-scale controlled experiments.
Agent-Based Modeling: a new approach
Another approach is to describe those individual interactions themselves, along the lines of the natural sciences, and use simulations of lots of digital individuals, for example, to assess whether then indeed those emergent trends in society are reproduced. This is the field of computational social science. The digital individuals are often referred to with the word “agents,” to indicate that they are, of course, simplified versions of a small selection of choices and actions that individuals (or households, or firms) can make.
What is known in the natural sciences as many-particle simulations, frequently in use in statistical physics and astrophysics, among others, the term “agent-based modeling” (ABM) is used in the social and economic sciences.
Applications and social impact
There is great societal interest in doing this kind of ABM because of the major societal challenges of inequality, climate adaptation, and public health. Human behavior and the feedbacks to these challenges need to be better understood in order to assess, for example, what policies will actually be effective. For example, consider that human behavior brings about climate change and that climate change also forces changes in behavior.
One example, from the field of public health, is how the spread of covid-19 can be related to patterns of human contact and mobility. An ABM has been developed for this purpose for infectious diseases.
It is now also clear that those models gain in quality by adding an agent for everyone in the Netherlands to the model: and thus modeling 18+ million agents and all their interactions. The field of economics and the emergence of economic crises also uses ABM. Again, it is the interactions between, and choices of, parties active in financial markets that themselves push an economy as a whole toward a critical state.
Many studies are also possible and needed on all forms of connections, and their impact on social inequality and segregation, using large-scale person networks. This research even without simulations already requires a lot of large-scale computation: the contact matrix of this persons network has 18 million x 18 million elements.
If this research can ultimately lead to better policies, because the impact of those policies can be calculated more objectively and in more detail, in a collaboration between academia, CBS and the planning agencies, and ministries, and of course other social and commercial organizations.
