Realizing the Potential of Agent-Based Social Simulation
Computer simulation is an established method to investigate the behavior of a system under certain conditions. When real-world experiments are too costly, time-consuming, impracticable, or impossible, simulation provides an alternative approach for controlled and systematic investigations. The insights from such experiments cannot only be used by scientists but more importantly also by decision-makers to consider the advantages and disadvantages of possible policies or actions before they are implemented.
When investigating the effects and success of different policies, a major factor that needs to be considered is the behavior of the individuals that are affected by the policy. Agent-Based Social Simulation (ABSS) is a simulation paradigm where AI is used to imitate human behavior and social interaction in a realistic way, making it particularly well-suited to investigate social phenomena. In practice, however, ABSS is mostly used by scientists and has not yet been established as a method in policy-making processes. Hence, we argue that the full potential of ABSS yet has not been realized in terms of providing support for societal decision-makers.
The goal of this project is to facilitate the use of ABSS in a twofold way. One the one hand, we will closely collaborate with policy makers to identify barriers that prevent the use of ABSS to support decision making. Potential reasons might include difficulties in selecting and using an appropriate model, unrealistic assumptions that are made by the models, or unawareness of the existence of ABSS. Through different case studies with societal policy makers from different sectors, we want to better understand the requirements that ABSS must fulfill, to provide the support that is required.
On the other hand, we believe that more sophisticated and trustworthy models are needed to optimally facilitate policy making. One issue is, for instance, the scalability of ABSS. Models often works great when the number of simulated individuals is in the thousands, yet, it is challenging to simulate millions on individuals such as the population of an entire state. Moreover, models of human behavior used in current ABSS are rather homogenous and do often not consider the actual variations in populations, e.g., age, gender, ethnicity, cultural characteristics, and habits.
The proposed project concerns removing the barriers that currently hinder ABSS to be used broadly by societal decision-makers and analysts. This includes the investigation of how the strengths of ABSS models can be combined with other models for investigating social behavior to novel and more mature models that allow for comprehensive and confident simulations of societal phenomena. The project will contribute to creating synergies between different disciplines and paradigms, and the generation of more sound simulation results.
Assistant Professor, Malmö University