How to Improve Network Science: the Potential of (Empirically Calibrated and Validated) Agent-Based Modelling

This article argues that the potential of Agent-Based Modelling (the capability for empirical justification of computer programmes representing social processes as dynamically unfolding individual cognition, action and interaction to reveal emerging aggregate outcomes) is not yet fully realised in the scientific study of social networks. By critically analysing several existing studies, it shows why the technique’s distinctive methodology (involving empirical calibration and validation) is just as important to its scientific contribution as its novel technical capabilities. The article shows the advantages of Agent-Based Models following this methodology and distinguishes these clearly from the implications of apparently similar techniques (like actor-based approaches). The article also discusses the limitations of existing Agent-Based Modelling applied to social networks, enabling the approach to make a more effective contribution to Network Science in future.