Elicitation of Probability Assessments
Engineering Honours Degree 2008
University of Adelaide
Decision making processes within the oil and gas industry are full of uncertainty, especially in exploration and appraisal activities. These uncertainties are generally reduced by further data collection, although there is always an element of interpretation and estimation required by oil and gas personnel. These uncertainty estimations are affected by cognitive biases [Welsh et al., 2005]. Of these biases, overconfidence is commonly present and is a focus of this study. Several previous studies, discussed herein, are aimed at overconfidence and the accuracy of uncertainty limits given for single confidence intervals.
This study investigates the accuracy of uncertainty estimations in the form of probability density function (PDF) elicitations. The motivation behind developing PDF elicitation methods, is that PDFs allow the interpretation of an infinite number of confidence intervals. There is also the possibility that PDF elicitation may somehow reduce overconfidence. Two different methods for PDF elicitations are tested, one based on initial most probable estimates, and the other based on initial maximum and minimum estimations. Approximately half of the sample had decision making training in the past twelve months, and therefore the effect of this training is also investigated.
The findings show a similar level of overconfidence in PDF elicitation methods to those found in uncertainty estimates for single confidence intervals [Welsh et al., 2005]. The comparison of PDF elicitation methods showed that the method based on minimum and maximum estimates produced a lesser degree of overconfidence. Comparisons also give reason to believe that decision making training has some effect in the reduction of overconfidence.