Statistics Which Will Lead Us to the Moon

Uncertainty and risk in decision-making, with application to the planning of NASA lunar space missions is the subject of investigation in the work Composite Multilinearity, Epistemic Uncertainty and Risk Achievement Worth by Emanuele Borgonovo (Department of Decision Sciences) and Curtis Lee Smith (Idaho National Laboratories, INL) to appear in the European Journal of Operational Research.

"This article was based upon work sponsored by an agency of the United States government", the usual disclaimer of US agencies, signals that this is the result of a research project between ELEUSI Bocconi and INL, one of the main US national labs. The project is part of the Faculty-Staff-Exchange program of INL, which has seen Bocconi as the first non-US University in the program for three consecutive years.

In modeling complex decision-making problems it is often necessary to distinguish between aleatory uncertainty, i.e., uncertainty which is intrinsic in the phenomenon under investigation, and epistemic uncertainty, i.e., uncertainty which is due to our lack of knowledge (or lack of data) and that can be reduced through further information. When the attention of the decision-maker needs to be pointed towards the most important elements of the problem, one needs quantitative indicators, called importance measures. These indicators must be consistent with the decision-maker's state of knowledge about the problem at hand.

Risk achievement worth is a number traditionally used to express how much the risk associated with the problem under investigation increases if a given event occurs. Traditionally, however, one considers only the aleatory part of the problem, overlooking epistemic uncertainty.

Borgonovo and Smith propose to remedy this shortcoming by introducing a new importance measure called epistemic risk achievement worth (ERAW). Epistemic means that this increase can now be assessed in a robust way, i.e., considering also the decision-maker's state of information about the problem at hand. In their work, Borgonovo and Smith also derive the conditions under which epistemic uncertainty does not matter in a decision-making problem. The conditions, however, reveal to be restrictive for realistic applications. The authors then apply the new importance measure to the risk analysis of a complex operational problem, namely, the design phase of a lunar space missions. The findings are then capable of conveying decision-makers in a robust way information about the elements of the problem on which to focus for most effectively preventing the occurrence of the undesired outcomes.