Validating Agent Interactions in Construct Against Empirical Communication Networks Using the Calibrated Grounding Technique

Validating a computational model is important as it establishes that the model has met its intended purpose of representing the system under study. In this paper, we perform a validation study on Construct, a multiagent network model for the coevolution of agents and the sociocultural environments that they inhabit. In particular, we focus on validating agent interactions produced by the model against empirical communication networks collected in real-world organizations. Validation is performed using our novel calibrated grounding technique. Results show that Construct can produce valid agent interactions. The benefits and implications of the study are discussed.