On the Formal Expression of Observational Data in Scientific Inference: Ambiguities and Confusions

摘要

As part of a programme to formulate scientific reasoning in terms of probabilistic inference, we investigate the crucial role of observations in conferring empirical content to theoretical models. Their expression within current frameworks is also examined, in the light of a canon of elementary common sense requirements. First we point out the most extended confusions regarding observational notions. Next we revise the ordinary conception of a datum as a realization of a variable and argue that conventional practice is insufficient to reflect the singularity of observed data, by failing to distinguish between contextual and factual evidences, and between actuality and potentiality. We conclude that the source of these difficulties is the inexactitude of formal languages with respect to observations.