Machines of War

The language of “programming” was fairly suffused with the military preoccupation with C3I in the 1940s, something all too readily forgotten, now that computers have osmosed into every comer of modem life. When the first digital computers were invented, designing a physical logic circuit was treated as part and parcel of the design of the algorithm it would execute, and instructions were frequently patched with phone plugs rather than “written” in symbols. As Dantzig explained with his coauthor Wood, the distinction between the planning of an elaborate calculation and the planning of a social structure was likewise elided in the development of “linear programming.” “It’s purpose is close control of an organization …. Programming … may be defined as the construction of a schedule of actions by means of which an economy, organization or other complex of activities may move from one defined state to another” (in Koopmans, 1951, p. 15). The evocative language of the finite state automata was no rhetorical flourish here, as elsewhere. Yet there was another attractive feature of the word “programming,” for which many participants at RAND shared an insider’s appreciation. It was not at all obvious just how long the Air Force would tum a blind eye to the amount of “pure” or playful mathematical research that was being funded at RAND; and there was the converse problem that one could not be too explicit in the 1950s as to the breadth of purview of ambitions to “plan” any economic process.

The beauty of that protean word “programming” was not only that it evoked the computer, but it would prove a convenient euphemism for any potentially controversial doctrine concerning control and power that had to be encapsulated with a certain opacity. 38 It became the most effective of the serried ranks of euphemisms for political phenomena at RAND, such as “organizational theory,” “systems analysis,” “logistics support,” and “decision analysis” (Fisher & Walker, 1994). The language of programming, and its polymorphously promiscuous referents, constituted a bridge spanning the concerns of the economists, the logicians, the engineers, and the operations researchers. This was not the first time the computer had served such a function, and neither would it be the last. Not only did the mathematics of linear programming resemble that of von Neumann’s game theory; the very act of programming a problem freighted with it many of the resonances of playing a game. As Paul Edwards reminds us, “all computer programming, in any language, is gamelike” (1996, p. 170). It was the structured manipulation of abstract symbols according to well-defined rules in order to arrive at preconceived outputs. Validity in such a reference frame comes to be assessed primarily in terms of internal coherence and consistency, and doing the mathematics came to 100m increasingly large in significance relative to demonstrated competence in the task at hand. Strategies could be treated as subroutines, and, given the emphasis on mixed strategies in game theory, statistical inference came to be conflated with the mixing and matching of algorithms. Hence the language of “programming” fostered an atmosphere within which Cowles could initially appear noncommittal about whether it was concertedly engaged in actually doing game theory, or just baroque optimization, or perhaps something else equally machinelike.

Mirowski, Philip (2001–12–03). Machine Dreams: Economics Becomes a Cyborg Science (pp. 259–261). Cambridge University Press. Kindle Edition.