On the Verisimilitude of Tendency Hypotheses

Tendency hypotheses – T-hypotheses, for short –, such as “the individuals of the kind Y tend to be X”, are used within several empirical sciences and play an important role in some of them, for instance in social sciences. However, so far T-hypotheses have received little or no attention by philosophers of science and statisticians.1 An exception is the work made in the seventies of the past century by the statisticians and social scientists David K. Hildebrand, James D. Laing, and Howard Rosenthal who worked out – under the label of prediction logic –, an interesting approach to the analysis of T-hypotheses.2 In this paper our main goal is the introduction of appropriate measures for the verisimilitude of T-hypotheses.3 Our verisimilitude measures will be defined in terms of the feature contrast (FC-) measures of similarity proposed by the cognitive scientist Amos Tverski (1977). We shall proceed as follows. In Section 1, Tverski’s FC-measures of similarity for binary features are illustrated and suitably extended to quantitative features. Afterwards, such measures are applied in the definition of appropriate measures for the verisimilitude of universal and statistical hypotheses (Section 2) and T-hypotheses (Section 3).