Understanding the Social Contagion Effect of Safety Violations within a Construction Crew: A Hybrid Approach Using System Dynamics and Agent-Based Modeling

Abstract

Previous research has recognized the importance of eliminating safety violations in the context of a social group. However, the social contagion effect of safety violations within a construction crew has not been sufficiently understood. To address this deficiency, this research aims to develop a hybrid simulation approach to look into the cognitive, social, and organizational aspects that can determine the social contagion effect of safety violations within a construction crew. The hybrid approach integrates System Dynamics (SD) and Agent-based Modeling (ABM) to better represent the real world. Our findings show that different interventions should be employed for different work environments. Specifically, social interactions play a critical role at the modest hazard levels because workers in this situation may encounter more ambiguity or uncertainty. Interventions related to decreasing the contagion probability and the safety–productivity tradeoff should be given priority. For the low hazard situation, highly intensive management strategies are required before the occurrence of injuries or accidents. In contrast, for the high hazard situation, highly intensive proactive safety strategies should be supplemented by other interventions (e.g., a high safety goal) to further control safety violations. Therefore, this research provides a practical framework to examine how specific accident prevention measures, which interact with workers or environmental characteristics (i.e., the hazard level), can influence the social contagion effect of safety violations.

Understanding the Sociocognitive Process of Construction Workers’ Unsafe Behaviors: An Agent-Based Modeling Approach

“…This research develops an agent-based modeling approach to understand the sociocognitive process of construction workers' unsafe behavior. Well-established theories (e.g., cognitive psychology [35] and social learning [82]), model validation test (e.g., [39,60,72]), and related research on managers, foreman, and coworkers (e.g., [23,24,41]) are used during the process. The model was established using MATLAB R2015b software (The MathWorks, Natick, MA, USA), relevant empirical data were used as initial values, and the validation of the model was tested.…”

“…Lastly, the distribution of the actual risk level for the baseline model is set at a modest level by a triangular distribution of (0.1, 0.5, 0.9). The lowest value is 0.1 and the highest value is 0.9 to avoid extreme low and high conditions [72]. Triangular distributions are used to simulate risk in previous research [72].…”

“…The lowest value is 0.1 and the highest value is 0.9 to avoid extreme low and high conditions [72]. Triangular distributions are used to simulate risk in previous research [72].…”

A Discourse on the Incorporation of Organizational Factors into Probabilistic Risk Assessment: Key Questions and Categorical Review

“…Hybrid/multimethod techniques refer to the cases, where a combination of two different types of methods is used to get benefit from diverse modeling technique capabilities. For example, some studies (e.g., (Mohaghegh, 2010a)) used a combination of SD and BBN which is a temporal method, and another study (e.g., (Liang, Lin, & Zhang, 2018)) integrated ABM and SD, which is a spatiotemporal method.…”

“…From the combinatory viewpoint, hybrid/multimethod techniques are better candidates as they allow the modeler to use different modeling techniques for different aspects of organizational performance. For example, the hybrid category of SD and ABM (Category 5.2.1 in (Justin ) enables the combination of individual (i.e., bottom-up) and system-level (i.e., top-down) dynamics, where individual processes can change a system state, alter system information, and in turn affect individual agents in a continuous cycle of information exchange (Liang et al, 2018). Hybrid/Multimethod modeling techniques allow for the combination of multiple theoretical perspectives through the integration of probabilistic, rule-based behavior, and equation-based modeling techniques.…”