VALIDATION - THE ELUDING DEFINITION

The term ''validation'' is featured prominently in the literature on r adioactive high-level waste (HLW) disposal. There exists, however, no unique definition of ''validation'' although it is generally understoo d to be related to model testing using experiments. This paper reviews the several definitions of ''validation'' and proposes their categori zation into three main classes. The first class links validation to th e goal of predicting the physical world as faithfully as possible. Thi s view has been criticized as being unattainable and, in any event, un suitable for setting gods for the safety analyses. Other definitions ( Class 2) are strictly operational, and associate validation only to si tuations where the models can be tested against observational data. In this view, the decision to retain a predictive model for use in safet y assessments does not belong to the remit of ''validation''. The thir d class of definitions focuses, instead, ''validation'' on the quality of the decision-making process, which shifts the debate from validati on in the observational sense to ''reasonable assurance'' and ''confid ence building''. In this third view, one cannot determine when a model or a suite of models are actually ''validated''. The lack of consensu s on the meaning of ''validation'' is linked to its short history as a technical term. The first technical definition dates from the mid-fif ties. The term was adopted thereafter in the computer field and elevat ed to its present status by the computer revolution of the seventies a nd early eighties. The term has made its appearance in some HLW safety standards only in the late eighties and it is virtually unmentioned i n the low-level waste standards for disposal safety. The continued inf ormal use of the term ''validation'' in the field of HLW disposal can become cause for endless speculations. The paper proposes either aband oning the use of this term or agreeing to a common definition.