Experience in expert estimation of probabilities and correlations for rail line construction

To counteract cost escalation and delays in infrastructure construction, an estimation model has been developed that simulates project uncertainty with probability distributions, correlations, and disruptive events. When applied to a new rail line project, three model inputs were not available from the designers: time distributions, cost correlations, and probability of occurrence and cost/time impacts of disruptive events. Experts in the field of tunnel-, viaduct-, and earthwork construction were asked to estimate the unavailable information. This paper presents a brief overview of biases that occur when experts estimate probabilities, the calibration steps taken to prevent the experts from being affected by such biases, and example questions from the questionnaire, followed by an in-depth discussion of findings. Unexpectedly, the experts followed radically different estimation techniques: the tunnel expert recalled values from past projects, the viaduct expert calculated the required values, while the earthwork expert estimated the mode value, and from it, the extreme values. Depending on estimation technique, a bias influences or cannot influence an expert. It was also observed that experts not familiar with probability theory encountered major difficulties in correlation estimation, despite concept explanation and visualization. Thus, for a successful correlation estimation it is paramount that experts be acquainted with probability theory in addition to being knowledgeable in their fields.