The agent-based model canvas: A modeling lingua franca for computational social science

Advancement in computational social science requires the expertise of social, computational, and data scientists. The field thrives due to the diversity of backgrounds and skills that these researchers bring from their native disciplines. Researchers also bring diversity in terminology that creates challenges in maintaining the effective communication necessary to design and execute complex research projects. We posit that a tool is needed to support communication and coordination across the members of research teams in computational social science. We propose a version of that tool: the Agent-Based Model Canvas. The Canvas is composed of nine basic building blocks that provide a framework for social theory design, testing, improvement, codification, and discussion in a truly interdisciplinary environment. The purpose of the Canvas is to assist theory discovery and refinement in an iterative fashion. The first version of the Canvas will be a starting point to list untested hypotheses and rough ideas about the elements of the model. As this iterative development process continues, the hypotheses are revised based on data analysis and emerging insights. This process progressively leads to refined descriptions that are machine executable, testable hypotheses, and finally, fully tested theories.