'The problem itself persists': Problems as Missing Links between Concepts and Theories in Canguilhem's Historical Epistemology

Abstract

If one was to register the specific terminological use of the concept of the 'problem' in 20th century philosophy of the sciences and of scientific practice, it is not without a soupçon of irony that one would have to compile such an inventory. Without a doubt, or so Warren Weaver claimed in 1948, a formalised mathematics of the kind that had propelled the regime of stochastics and statistical mechanics about half a century earlier would hold the key to unlock so-called 'problems of disorganized complexity' (Weaver 1948: 538), that is to say logical situations involving a vast amount of variables, all of which display 'individually erratic' (ibid) characteristics. Conundrums of that type, Weaver argued, will turn out to be more complex to manage than 'two-variable' (ibid: 537) problems-which conveniently correspond to the binary system of the mathematics of mechanics-but it is through the recurrence of 'certain orderly and analyzable average properties' (ibid: 538) that, despite the numeric range and largely random interplay of the variables, predictions as to their standard distribution will hold true 'with increasing precision' (ibid). The reason for the relative controllability of such a system of randomly interconnected, yet homogeneous, elements is the 'disorganized complexity' of the ensemble, that is the absence of an internal order-an organised configuration of the components that would be irreducible to a recurring pattern of averages. As opposed to these two classes of problem constellations, 'problems of simplicity' (which can be addressed by the mathematics of mechanics) and 'problems of disorganized complexity' (mastered by stochastics), then, Weaver identifies a third type of problem that displays