Modelling socio-technical risks of industrial megaprojects using Bayesian Networks and reference classes

Current processes in definition, estimation, and financing of large resource projects often exclude relevant sources of socio-technical risks with overarching effects. This paper proposes an expandable Object-Oriented Bayesian Network (OOBN) to incorporate the effects and dependencies of project risk factors and outcome variables as related to cost. The model can be trained based on collected data, expert knowledge for previously unmeasured sources of socio-technical risks, or some combinations thereof. The model is linked to a reference class of mining projects to position the project at hand within the reference class and to distinguish various extents of cost overrun. Three rounds of interviews were conducted with project expert cohorts to identify socio-technical risk factors, establish their influence weights, and verify specific projects' results. The methodology helps overcome various biases within project estimation processes by combining both singular-evidence of the project at hand (i.e., inside view) and distributional-evidence of the peer reference class (i.e., outside view).