Abstract (EN)

Building a composite index (CI) is a delicate work implying a series of decisions which consequences will determine the scope of the tool. This is especially true for the very first steps of this construction: the definition of the theoretical framework and the selection of relevant indicators, usually referred to as the construction of the model. However, the literature seems to focus more on the other aspects, quite often eluding those first critical steps or considering it self-evident. How to ensure the soundness of those tools when the values that they convey are not even clearly expressed? How to ensure the completeness of the model according to the reality that it ought to synthetise when it's a highly complex or abstract one? Each of these steps involves hypotheses and presuppositions that must be explained and evaluated with care. In this article, we seek to identify a coherent method that will fit that very purpose. Our primary hypothesis, following recent work, is that conceiving CIs as potential decision problems is a relevant approach. It allows to solicit the Multi Criteria Decision Aiding theory, that is rich in modelling methodologies-such as Ralph Keeeney's Value Focused Thinking. Using a real-life application, the construction the Prison Life Index,we try to show the relevance of those primary hypothesis by displaying the model of the index and summarizing the 1 Social Indicators Research 2 CONTENTS stage of its construction. This adapted methodology can guide index builders in the determination of a model that comprehensively fits its intended purpose and that isn't overly dependent on available data.