L'enquête en gestion de la qualité dans l'industrie: une approche sémiotique

Industrial quality control is an interesting domain for studying knowledge and knowing because it involves a strong theoretical dimension (mathematical statistics) together with industrial work practices. Methods for statistical quality control have been developed in the years 1920s to address the problems of mass production which made impossible the inspection of each product individually. To make those mathematical methods usable by ordinary workers on the shopfloor, graphical tools have been developed ("Shewhart control charts"), which translate formal statements into metaphorical representations drawn from ordinary life. Thus, "monitoring an industrial process" becomes, on the chart, "keeping a point within two limiting lines", the point being representative of a sample drawn according to a pre-defined procedure. This graphical tool is an artefact which endows the mathematical theory with a social life. These statistical procedures are designed to be "optimal" from an economical point of view, for example by minimizing the cost of sampling to acquire a given amount of information, or by evaluating and balancing the risks of making a bad decision of type I as opposed to a bad decision of type II in a statistical test. These procedures thus appear to bear all the intelligence that is needed in the situation. We might thus ask what amount of intelligence is left to the worker's care in practice. An answer resides in another aspect of quality control which I call "the inquiry" : it consists in finding the causes and remedies for quality defects observed during production. This aspect does not usually pertain to statistical theory, at least to the theory of sampling and hypothesis testing. Obviously, there is a lack of conceptual articulation between these two aspects of quality control. In this paper, I propose to conceptualize the inquiry by means of semiotics, relying on work by Charles S. Peirce and Umberto Eco. In case of quality defects of significative importance according to statistical criteria, the control chart gives a signal which is an index, in Peircean terms. The inquiry can be conceptualized as a process of semiosis which enriches this index sign into full triadic signs. The semiosis develops mainly trough abduction, and we find useful work on this topic by Eco. As an example, this conceptualization helps to clarify the meaning of a management tool widely used in industrial operations management, known as the "Deming Cycle".