| Summary: | Social phenomena represent a complex subject for scientific study due to the inherent complexities of human behavior, societal structures, and community dynamics. This interplay can be multifaceted and exhibit stochastic properties, rendering their manual reproduction or replication through specific models costly, intricate, and sometimes impracticable. Agent-based simulation is a way to explore global patterns, such as community dynamics, particularly in scenarios where interactions between individuals are reasonably well-defined. However, a significant drawback of contemporary agent-based models is how adjustments are made in light of new information on the phenomenon under investigation. Implementing adjustments in complex models can prove to be resource-intensive or even infeasible, posing significant challenges to their continued applicability. As such, there is a necessity for extensible and adaptive modeling approaches that can accommodate new information and maintain relevance in the face of evolving understanding of complex systems. While toolkits have emerged to simplify the implementation of agent-based models with numerous benefits, they present the same limitations as current implementations and particular challenges associated with simulating complex models. In this thesis, we introduce a new methodology for employing systems that cater to the simulation of highly complex and changing social phenomena, where the representation of domain-specific concepts is achieved through the use of ontologies input into a generalized agent-based model framework to run a simulation. To test the utility and generality of the method, we employ the use-case of substance use in adolescents and young children, a study with longitudinal data collected in Iceland, by producing scenarios necessary to simulate the subject but sufficiently general and applicable to other social systems. Our proposed approach, which emphasizes modularity and flexibility, reveals a promising capability to encapsulate the discussed scenarios ...
|
|---|