Enhancing the Credibility of Agent-Based Model for the Study of Workers' Group Behavior by Comparing Simulation Data with Survey Data

摘要

Construction workers' behavior is an important factor of productivity and safety on a job site. Worker behavior is not only determined by individual's inherent characteristics, but also largely shaped by contexts which include the social interactions within a project. As a result, construction workers' behavior on a job site is under the influence of the social norms and work culture of their workgroup. However, our current understanding of the social aspect of worker behavior is limited. To expand our understanding of workers' social behavior and its impact on project performance, researchers have begun using an agent-based modeling approach because it allows researchers to observe the complex group-level behavior emerging from individuals' interactions. One of the important issues in using agent-based models for organizational research is enhancing the credibility of model predictions. With this background in mind, the objective of this paper is to propose a methodology that can help enhance the credibility of an agent-based model for the study of workers' social behavior using survey data. Specifically, it is suggested in this paper that the perceptual/attitudinal/behavioral data from surveys and the corresponding data from simulations be transformed into categorical data, and then compared to each other to show a quantitative agreement of the model behavior with empirical data. The proposed methodology is illustrated by an example of construction workers' absenteeism research that we have conducted. With the results of this research, it is argued that demonstrating the correspondence of simulation data to survey data using a data categorization method is an effective means to enhance the credibility of an agent-based model.