Definition of Term: Elicitation

Probability distributions are needed to express uncertainty about unknown quantities in various contexts. In the MUCM toolkit, there are two principal areas where such probability distributions may be required. One is to specify prior knowledge about hyperparameters associated with building a Gaussian process emulator, while the other is to formulate uncertainty about inputs to the simulator (for instance, for the purposes of uncertainty or sensitivity analyses). These probability distributions may be created in various ways, but the most basic is to represent the knowledge/beliefs of a relevant expert.

The process of formulating an expert’s knowledge about an uncertain quantity as a probability distribution is called elicitation.