iEMSs 2022 Conference - Brussels, Belgium

iEMSs 2022 - Session C.4

Reusable building blocks for agent-based models of social-environmental systems.

Stream    : C - Computational Methods, Workflows, Informatics and Integrated Systems in Environmental Modelling

Session Leader: Uta Berger, TU Dresden; Tatiana Filatova, Delft University of Technology; Volker Grimm, Helmholtz Centre for Environmental Resarch-UFZ

Agent-based models (ABM) of socio-environmental systems (SES), representing the behavior of organisms, human actors, or institutions, are usually fully developed from scratch. This is inefficient and leads to incoherent model designs, with multiple variations of how the same types of agents’ decisions are coded. Making ABMs open access is a major step towards transparency and reusability, but in practice long, complex, case- or problem-dependent computational codes are rarely reused. This hinders development of robust and effective SES ABM applications, that are well aligned with micro-level behavioral, social and ecological theories at the agent level and with the general system level theories. It is therefore important to adopt the strategy from other modelling communities to build a repository of reusable building blocks (RBB). This would help agent-based modelers to focus their energy and creativity on missing parts, and scientists without computer science background to efficiently assemble powerful simulation SES ABMs and concentrate on their particular research questions. RBBs are submodels of certain behaviors and processes which are likely to be important for many ABMs in a given discipline or application, for example foraging in ecology, farming decisions in land use models, or households’ decisions to install solar panels. An open access library of commonly used RBBs could rely on a solid theoretical microfoundations for agents rules of actions and interactions, be tested in different empirical SES contexts, and improve evolutionary over time. Established RBBs would have known properties, be well-tested by the community, and be reusable in different contexts. For this, they would need to be uploaded to repositories providing: source code, written model description, specification of required context, executable demonstration, and reports of tests and example applications. Also, RBB standards should be developed. For this session, we welcome contributions of specific candidate RBBs and of methodological considerations regarding the modular development of ABMs, including standardization of uploads, web-based resources, version control, incentives for uploading as well as pros and cons of using RBBs for theory development and advancing ABM applications for SES.