Structural Uncertainties in Numerical Induction Models

This report delineates a number of ways in which the results of numerical induction models, which aggregate lower level measures into meta-measures for decision making, can be unnecessarily compromised. Examples of numerical induction models include complex models for performance evaluation, measures of effectiveness synthesis, and for strategic decision analysis. A framework is proposed for identifying different types of modelling uncertainty that may be present and several of these uncertainties are discussed in detail. Some popular decision analysis techniques are also analysed highlighting any features that may introduce unnecessary uncertainty into the results. The purpose of describing these potential pitfalls is to reduce the structural uncertainty forms that may be unwittingly added to the uncertainties that already exist in the input information leading to outputs that are more meaningful. More meaningful outputs should then naturally result in improved decisions when such models are applied to Defence problems.