noble machine design

3 SOCIAL CHOICE IN MACHINE DESIGN D.F. NOBLE D.F. Noble 'Social choice in machine design', in A. Zimbalist (ed.), Case Studies on the Labor Process, Monthly Review Press, New York, 1979, pp. 18-50. Introduction Almost everyone would agree that the technology of production and the social relations of production are somehow related. The explanation of this relationship often takes the form of a more or less "hard" technological determinism: Technology is the independent variable which effects changes in social relations; it has its own immanent dynamic and unilinear path of development. Further. it is an irreducible first cause from which social effects automatically follow. These effects are commonly called its "social impact." Social analysts have recently begun to acknowledge that the technology and the social changes it seems to bring about are in reality interdependent, and it has become fashionable to talk about the dialectic be,ween the forces of production and social relations. Nevertheless. most studies of production continue to focus primarily on the ways in which technology affects social relations and there is precicus liu:e effort made to show precisely how technology reflects them. That is, although grantsmanship now demands that people refer to the mutual dependenct.· of technology and society, and although socialists and other radicals now take it for granted that technological developm(;.nt is socially determined, there remains very little con<:rete, historical analysis that demonstrates the validity of the position. The present essay. a case history of the design, deployment, and actual use of automatically controlled machine tools. is meant to be a step in that direction. Elsewhere I have tried to show that technology is not an autonomous force impinging upon human affairs from the "outside," but is the product of a social process, a historically specific activity carried on by some people, and not others, for partinilar purposes (Noble 1977). Technology thus does not develop in a unilincar fashion; there is always a range of possibilities or alternatives that are delimited over time-as some 103 () ., are selected and others denied-by the social choices of those with the power to choose, choices which reflect their intentions, ideology, social po",tion, and relations with other people in society. ! n short, technology bears the social "imprint" of its authors. It follows that "social impacts" issue not so much from the technology of production as from the social choices that technology embodies. Technology, i:hen, is not an irredudble first cause; its social effects follow from the social causes that brought it into being: behind the technology that affects· social relations lie the very same social relations. Little wonder, then. that the technology usually tends to reinforce rather than subvert those relations. Here I want to render this abstract argument concrete by examining a particular technology. Moreover. I want to go a step further and show that the relationship between cause and effect is never automatic-whether the cause is the technology or the social choices that lie behind it-but is always mediated by a complex process whose outcome depends, in the last analysis, upon the relative strengths of the parties involved. As a result, actual effects are often n.ot consonant with the expectations implicit in the original designs. The technology of production is thus twice determined by the social relations of production : first. it is designed and deployed according to the ideology and social power of those who make such decisions; and second, its actual use in production is determined by the realities of t!'le shop-floor struggles between classes. This essay is divided into six parts. A description and brief history of the technology involved is followed by a two-part section on social choice in design that discusses both the horizontal relations of production (between firms) and the vertical relations of production (between capital and labor). The fourth part examines social choice in the deployment of technology and the fifth looks at shop-floor realities where this technology is being used in the United States today. In the last part some alternative realities, with different social relations, are described. The Technology: Automatically Controlled Machine Tools The focus here is numerically controlled machine t~ls, a particular production technology of relatively recent vintage. 104 According to many observers, the advent of this new technology has produced romething of a revolution in manufacturing, a revolution which, among other things, is leading to increased concentration in the metalworking industry and to a reorganization of the production process in the direction of greater managerial control. These changes in the horizontal and ver~ical relations of production are seen to follow logically and inevitably from the introduction of the new technology. "We will see some companies die, but I think we will see other companies grow very rapidly," a sanguine president of Data Systems Corporation opined (Stephanz 1971 ). Less sanguine are the owners of the vast majority of the smaller metalworking firms which, in 197 l, constituted 83 percent of the industry; they have been less able to adopt the new tec1'nology be.cause of the very high initial expense of the hardware, and the overhead and difficulties associated with the software (ibid). In addition, within the larger, better endowed shops, where · the technology has been introduced, another change in social relations has been taking place. Earl Lundgren, a sociologist who surveyed these shops in the late 1960s, observed a dramatic transfer of planning and-control from the shop floor to the office ( 1969). For the technological determinist, the story is pretty much told: numerical control leads to industrial concentration and greater managerial control over the production process. The social analyst, having identified the cause, has only to describe the inevitable effects. For the critical observer, however, the problem has merely been defined. This new technology was developed under the auspices of management within the large metalworking firms. Is it just a coincidence that the technology tends to strengthen the market position of these firms and enhance managerial authority in the shop? Why did this new technology take the form that it did, a form which seems to have rendered it accessible only to some firms, and why only this technology? Is there any other way to automate machine tools. a technology, for example, which would lend itself less to managerial control? To answer these questions, let us take a closer look at the technolugy. A machire tool (for instance, a lathe or milling machine) is a machine used to cut away surplus material from a piece of metal in order to produce a part wi_th the desired shape, size, and finish. Machine tools are really the guts of machine-based industry because they are the means whereby all machinery, including 105 1 {)) the machine tools themselves, are made. The machine tool has traditionally been operated by a machinist who transmits his skill and purpose to the machine by means of cranks, levers, and handles. Feedback is achieved through hands, ears, and eyes. Throughout the nineteenth century, technical advances in machining developed by innovative machinists built some intelligence into the machine tools themselves-automatic feeds, stops, throw-out dogs, mechanical cams-making them partially "self-acting." These mechanical devices relieved the machinist of certain manual tasks, but he retained controi over the operation of the machine. Together with elaborate tooling-fixtures for holding the workpiece in the proper cutting position and jigs for guiding the_ path of the cutting tool-these design innovations made it possible for less skilled operators to use the machines to cut parts after they had been properly "set up" by more skilled men;* but the source of the intelligence was still the skilled machinist on the floor. The 1930s and 1940s saw the development of tracer technology. Here patterns, or templates, were traced by a hydraulic or electronic sensing device which then conveyed the information to a cutting tool which reproduced the pattern in the workpie~~- Tracer technology made possible elaborate contour cutting, but • The use of jigs and fixtures in metalworking dates back to the early nineteenth century and was the heart of interchangeable parts manufacture, as :\ferritt Roe Smith has shown (1976). Eventually, in the closing decades of the century, the "toolmaker" as such became a specialized trade, distinguished from the machinist. The new function was a product of modern management, which aimed to shift the locus of skill and control from the production floor, and the operators. to the toolroom. But however much the new tools allowed manage· ment to employ less skilled, and thus cheaper. machine operators, they were nevenheless very expensive to manufacture and store and they lent to manufacture a heavy burden of inflexibility, shortcomings which one Tayiorite, Sterling Bunnell. warned about as early as 1914 (cited in David Montgomery. unpublished ms.). The cost•savings that resulted from the use of cheaper labor were thus partially offset by the expense of tooling. Numerical control, as we will see, was developed in "Part to eliminate the cost and inflexibility of jigs and fixtures and, equally important, to take skill, and the control ofit, off the floor altogether. Here again, however, the expense of the solution was equal to or greater than the problem. It is interesting to note that in both cases expensive new technologies were introduced to make it possible to hire cheaper labor, and the tab for the conversion was picked up by rhc: state-the Ordnance Department in the early nineteenth century, the departwents of the army and navy in World War I. and the air force in the second half of the twentieth century. BEST COPY AVAILARlE 106 it was only a partial form of automation: for instance, different templates were needed for different surfaces on the same workpiece. With the war-spurred development of a whole host of new sensing and measuring devices, as well as precision servomotors which made possible the accurate control of mechanical motion, people began to think about tte possibility of completely automating contour machining. Automating a machine tool is different from automatin_g, say. automotive manufacturing equipment, which is single-pu'rpose, fixed automation, and cost-effective only if high demand makes possible a high product volume. Machine tools are general purpose, versatile machines. used primarily for small batch, low volume producti_on of parts._ The challenge .of automating machine tools, then; was to render them self-acting while retaining their versatility. The solution was to develop a mechanism that translated electrical signals into machine motion and a medium (film, lines on paper, magnetic or punched paper tape, punched cards) on which the information could be stored and from which the signals could be reproduced. The automating of machine tools, then, involves two separate processes. You need tape-reading ~nd machine controls, a means of transmitting information from the medium to the machine to make the tables and cutting tool move as desired, and you need a means of getting the information on the medium, the tape, in the first place. The real challenge was the latter. Machine controls were just another step in a known direction, an extension of gunfire control technology developed during the war. The tape preparation was something new. The first viable solution was "record-playback,'' a system developed in 1946-194 7 by General Electric, Gisholt, and a few smaller firms.* It involved having a machinist make a part while- the motions of the machine under his command were recorded on magnetic ~ape. After the first piece was made, identical parts could be made automatically by playing back the tape and reproducing the machine motions. John Diebold, a management • The discussion of the record-pla}·back technology is based upon extensive interviews and correspondence with the engineers who participated in the projects at ~nc:ral Electric (S.:henectaciy) and Gisholt (Madison. Wisconsin), and the trade journal and technical literature. 107 104 consultant and one of the first people to write about "flexible automation," heralded record-playback as "no small achievement . . . it means that automatic operation of machine tools is possible for the job shop-normally the last place in which anyone would expect even partial automath>n" (1952:88). But record-playback enjoyed only a brief existence, for reasons we shall explore. (It was nevertheless immortalized as the inspiration for Kurt Vonnegut's Player Piano. Vonnegut was a publicist at GE at the time and saw the record-playback lathe which he describes in the novel.) The second solution to the medium-preparation problem was "numerical control'' (N/C), a name coined by M1T engineers William Pease and James McDonough. Although some trace its history.back to the Jacquard loom of 1804, N/C was in fact of more recent vintage; the brainchild of John Parsons, an air force subcontractor in Michigan who manufactured rotor blades for Sikorski and "Bell helicopters. In 1949 Parsons successfully sold the air force on his ideas, and then co,1tracted o·ut most of the research work to the Servomechanisms Laboratory at MIT; three years later the first numerically controlled machine tool, a vertical milling machine, was demonstrated and widely publicized. Record-playback was, in reality, a multiplier of skill, simply a means of obtaining repeatability. The intelligence of production still came from the machinist who made the tape by producing the first part. Numerical control. however, was based upon an entirely different philosophy of manufacturing. The specifications for a part-the information contained in an engineering blueprint-are first broken down into a mat~ematical representation of the part, then into a mathematical description of the desired path of the cutting tool along up to five axes. and finally into hundreds or thousands of discrete instructions, translated for econ0my into a numerical code. which is reacl and translated into electrical signals for the machine controls. The N/C tape, in short, is a mea.ns of formally circumventing the role of the machinist as the source of the intelligence of production. This new approach to machining was heralded by the National Commission on Technology, Automation, and Economic Progress as "probably the most significant development in manufacturing since the introduction of the moving assembly line'' (Lynn et al. 1966:89). 10~ 108 Choice in Design: Horizontal Relations of Production This short history of the automation of machine tools describes the evolution of new technology as if it were simply a technical, and thus logical, development. Hence it tells us very little about why the technology tC'Ok the form that it did, why N/C was developed while record-playback was not, or why N/C as it was designed proved difficult for the metalworking industry as a whole to absorb. Answers to questions such as these require a closer look at the social context in which the N/C technology was developed. In this section we will look at the ways in which the design of the N/C technology reflected the horizontal relations of production, those between firms. In the following section. we will explore why N/C was chosen over record-playback by looking at the vertical relations of production, those between labor and management. . To begin with, we mus~ examine the nature of the machinetool industry itself. This tiny industry which produces capital goods for the nation's manufacturers is a boom or bust industry that is very sensitive to fluctuations in the business cycle, experiencing an exaggerated impact of good times-when everybody buys new equipment--and bad times-when nobody buys. Moreover, there is an emphasis on the production of "special" machines, essentiallr custom-made for users. These two factors explain much of the cost of 1nachine tools: manufacturers devote their attention to the requirements of the larger users so that they can cash in on · the demand for high-performance specialized machinery. which is very expensive due to high labor costs and the relatively inefficient low-volume production methods (see Rosenberg 1963; Wagoner 1968; Brown and Roser.berg 196 l; Melman 1959). The development of N/C exaggerated these tendencies. John Parsons conceived of the ne,,· ttchnology while trying to figure out a way of cutting the difficult contours of helicopter rotor blade templates to close tolerances; sir.c:e he was using a computer to calculate the points for drilling holes (which were then filed together to make the contour) he began to think of having the computer control the actual positioning of the drill itself. He extended this idea to three-axis milling when he examined the spedfication for a wing panel for a new combat fighter. The new highlOG 109 performance, high-$peed aircraft demanded a great deal of difficult and expensive machining to produce airfoils (wing surfaces, jet engine blades), integrally stiffened wing sections for greater tensile strength anci less weight, and variable thickness skins. Parsons took his idea, christened "Cardomatic'! after the IBM cards he used, to Wright Patterson Air Force Base and convinced people at the Air Material Command that the air force should underwrite the development of this potent new technology. When Parsons got the contract, he subcontracted with MIT's Servomechanism Laboratory, which had experience in g~nfire control systems.• Between the signing of the initial contract in 1949 and 1959, when the air force ceased its formal support for the development of software, the military spent at least $62 million on the research, development, and transfer of N/C. Up until 1953, the air force and MIT mounted a large campaign to interest !'Tlachine-tool builders and the aircraft industry in the new technology, but only one company, Giddings · and Lewis, was sufficiently interested to put their own money into it. Then, in 1955, N/C promoters succeeded in having the specifications in the Air Material Command budget allocation for the stockpiling of machine tools changed from tracercontrolled machines to N/C machines. At that time, the only fully t-i/C machine in existence was in the Servomechanism Lab. The air force undertook to pay for the purchase, installation, and maintenance of over 100 N/C machines in factories of prime subcontractors; the contractors, aircraft manufacturers, and their suppliers would also be paid to learn to use the new technology. In short, the air force created a market for N/C. Not surprisingly. machine-tool builders got into action, and research and development expenditure in the industry multiplied eightfold between 1951 and 1957. The point is that what made N/C possible-massive air force support-also helped determine the shape the technology would take. While criteria for the design of machinery normally includes cost to the user, here this was not a major consideration; machine-tool -builders were simply competing to meet performance and "competence" spedfications for government-funded • This brief history of the origins of ~IC is based upon interviews with Parsons and MIT personnel, as well as the use of Parsons' personal files and the project records of the Servomechani~m Laboratory. 10'/ 110 users in the aircraft industry. They had littie concern with cost effectiveness anci absolutely no incentive to produce less expensive machinery for the commercial market. But the development of the machinery itself is only part of the story; there was also the separate evolution of the software. Here. too, air force requirements dictated the shape of the technology. At the outset, no one fully appreciated the difficulty of getting the intelligence of production on tape, least of all the MIT engineers on the N/C project, few of whom had had any machining experience before beroming involved in the project. Although they were primarily control engineers and mathematicians, they had sufiicient hubris to believe that they could readily synthesize the skill of a machinist. It did not take them long to discover their error. Once it ,vas clear that tape preparation was the stumbling block to N/C's economic viability, programming became the major focus of the project. The first programs were prepared manually. a tedious. time-consumiug operation performed by graduate students, but thereafter efforts were made to enlist the aid of Whirlwind. :\HT's first digital computer. The earliest programs were essentially subrQutines for particular geometric surfaces which were compiled by an executi\'e program. In 1956. after ~I IT had received another air force contract for software de\'elopm~nt, a young engineer and mathematician named Douglas Ross came up with a new approach to programming. Rather than treating each separate problem with a separate subroutine, the new system, called APT (Automatically Programmed Tools), ,,·as essentially a skeleton program-a "svstematized solution," as it was called-for moving a cutting tool through space; this skeleton was to be "fleshed out" for every particular application. The APT system \\"as flexible and fundamental; equally importan~ it met air force specifications that the language must have a capacity for up to fiveaxis control. The air force loved APT because of its flexibility; it seemed to allow for rapid mobilization. for rapid design cha1·•~e. and for interchangeability between machines ,,·ithin a plant, b tween users and vendors. and between contractors and subco11tractors throughout the country (presumabl~ of "strategic importance" in case of enemy attack). With these ends in mind. the air force pushed for standardization of the APT system and the Air Material Command cooperated with the Aircraft Industries Assoriation Committee on Numerical Control to make APT the industry standard. the machine tool and control manufac111 lC'-' turers followed suit, developing "postprocessors" to adapt each particular system for use with APT. Before long the APT computer language had becomt: the industr1 standard, despite initial resistance within aircraft. <..:.>r;.1- pany plants. Many of these companies had developed their own languages to program their N/C equipment, and these in-house languages, while less flexible than APT, were nevertheless proven, relatively simple to use, and suited to the needs of the company. APT was something else entirely. For all its advantages--indeed. because of them-the APT system had decided disadvantages. The more fundamental a system is, the more cumbersome it is, and the more complex it is, the more skilled a programmer must be, and the bigger a computer must be to handle the larger amount of information. In addition, the greater the amount of information, the greater the chance for error. But initial resistance was O\'ercome by higher level management, who had come to believe it. necessary to learn how to use the ne\v systern "tor business _reasons" (cost-plus contracts with the air force). The eXC!USi\'e use of APT was enforced. Thus began what Douglas Ross himself has described as "the tremendous turmoil of practicalities of the APT system de- \ elopment"; the system remained ''erratic and unreliable," and a major headache for the aircraft industrr for a long time. The standardization of APT, at the behest of the air force. had two other interrelated consequences. First, it inhibited for a decade the development of alternati\·e. simpler languages. such as the !>trictly numerical language :-,.; CFOR\I (created by A. S. Thomas, Inc.), which might have rendered contour programming more accessibk to smaller shops. Second. it forced those who ventured into N/C into a dependence on those who controlled the development of APT,* on large computers and • The air force funded development of APT was centered initiallv at MIT. In l'.J61 tl::e effort was shifted to the Illinois Institute of Technology Research Institute (IITRI) where it has been carried on undt>r the direction of a consortium composed o(the air forl:e, the Aircraft Industries Association (AIA), and major manufacturers of machine tools and electronic controls. ~1embcrship in the consortium has always been exper,~ive, bevond the financial means of the vast majority of firms in the metalworking industry. APT system use, therefore, has tended to be restricted to those who enjoyed privileged access 10 information about the system's d·evelopment. Moreover, the APT system has been treated as proprietar>· information within user plants; programmers have had to sign out lOD 112 mathematic.tily sophis1-icated programmers. The aircraft companies, for all their headaches, could afford to grapple with APT because of the air force subsidv, but commercial users were not so lucky. Companies that wanted military contracts were compelled to adopt the APT system, and those who could not afford the system, with its training requirements, its computer demands, and its headaches, were thus deprived of government jobs. The point here is that the software system which became the de facto standard in industry had been designed with a user, the air fcrce, in mind. As Ross explained, "the universal factor throughout the design process is the economics involved. The advantage to be derived from a given aspect of the language must be balanced against the difficulties in incorporating that aspect into a complete and working system" (Ross 1978:13). APT served the air force and the aircraft industry well. but at the expense of less endO\\·ed competitors. Choice in Design: Vertical Relations of Production Thus far we hare t:1 lked ont-. about the form of i'\/C. its hardware and software, and how these reflected the horizontal relations of production. But what about the precursor to 'SIC. record-playback~ Here was a technology that was apparently perfectlv suited to the small shop: tapes could be prepared bY recoding the motions of a machine tool. guided by a machinist 01· a tracer template. without programmers, mathematics. languages. or computers.* Yet this technology was abandoned in favor of ~IC by the aircraft industry and b~· the control manfor manual~ .1110 ha\·e been forbidden from taking them home or talking about their foments with people outside the companv. •Technically. record•pla, hack was as rcliahk as :-,;1c. if not more so-since all the programming was done at the machine. errors could be eliminated during the programming process. before production began. Moreover. it could be used to reproduce parts to wirhin a tolerance of a thousandth of an inch. just like N/C. (It is a common mistake 10 assume that if an N/C control system generates discrete pulses corresponding to increments of half a thousandth. the machine can produce parts 10 within the same toleranfes. In reality. the limits of accuracy are set by the machine it, clf-not to mention the weather-rather than by the electrical signals.) 1.t0 113 ufacturers. Small firms never saw it. The Gisholt system, designed by Hans Trechsel to be fully accessible to machinists on the floor, was shelved once that company was bought by Giddings and Lewis, one of the major N/C manufacturers. The GE record-playback system was never really marketed since demonstrations of the system for potential customers in the machinetool and aircraft companies elicited little enthusiasm. Giddings and Lewis did in fact purchase a record-playback control for a large profile "skin mill" at Lockheed but switched over to a modified N/C System before regular production got underway. GE's magnetic tape control system, the most popular system in the 1950s and 1960s, \,·as initiallv described in sales literature as having a "record-playback optio'n," but mention of this feature soon disappeared from the manuals, even though the system retained the record-playback capacity.* Why was there so little interest in this tt:chnology? The answer to this question is complicated. First, air force performance spe{:ifications for four- and fi\"e-axis machining of complex part~, often out of difficult materials, were simply beyond the capacity of either record-playback or manual methods. In te1·ms of expected cost reductions, moreover, neither of these methods appeared to make possible as much of a reduction in the manufacturing and storage costs of jigs. fixtures, and templates as did N/C. Along the same lines. N/C also promised to reduce more significantly the labor costs for toolmakers, machinists, and patternmakers. And, of course, the very large air force subsidizatirn. of N/C technology lured most manufacturers and users to where the action was. Yet there were still other, less practical. reasons for the adoption of ~IC and the abandonment of record-playback. reasons that have more to do with the ideology of engineeri,~g than with economic calculations. However useful as a production technology. record-playback was considered quaint from the start. especially with the advent of N/C. N/C was always more _than a technology for cutting metals, especialiy in the eyes of its MIT designers, who knew little about metakutting: it wa~- a symbol of the computer age, of mathematical elegance, of power, order, and predictability, of continuous * This hbton· is based upon irucrviews with Hans Trechsel, designer ot Gishnh"s "' Factrol"' s~stem, and intcn·iews and correspondence with participating engineering and sales personnel at GE (Schenectady), as well as articles in various engineering and trade journals. 111 114 flow, of remote control, of the automatic factory . Recordplayback, on the other hand, however much it represented a ~ignificant advance on manu~.l methods, retained a vestige of traditional auman skills; as such, in the eyes of the future (and engineers always confuse the present and the future) it was obsolete. · The drive for total automation which N/C represented, like the drive to substitute capital for labor. is not always altogether rational. This is not to say that the profit motive is insignificant-hardly. But economic explanations are not the whole story, especially in cases where ample government financing renders cost-minimization less of an imperative. Here the ideology of control emerges most clearly as a motivating force, an ideology in which the distrust of the human agency is paramount, in which human judgment is construed as "human error." But this ideology is itself a reflection of something else: the reality of the capitalist mode of production. The distrust of human beings by engineers is a manifestation of capital's distrust of labor. The elimination of human error and uncertaintv is the engineering expression of capital's attempt to minimize'its dependence upon labor by increasing its control o,·er production. The ideologv of engineering, in short. mirrors the antagonistic social relations of capitalist production. Insofar as the design of machinery. like machine tools, is informed by this ideology. it reflects the social relations of production.* Here we will emphasize this aspect of the explanation-why N/C was developed and record-playback was not-primarily because it is the aspect most often left out of such stories. • It could be argued that control in the capitalist mode of production is not an independent factor (a manifrstation of class conflict). but merely a means to an enmomic end (the accumulation of capital). Technolo~· introduced to increase managerial control O\'er the work force and eliminate pacing is, in this view. introduced simply to increase profits. Such reductionism. which collapses control and dass quc:-stions into cconomistic ones. renders impossible any explanation of technological development in terms of social relations or am· careful distinction between productive technology which directly increases output per person-hour and technologv "·hich de>es so onlv indirectly b~· reducing worker resistance or restriction of output. Finally. it makes it hard to distinguish a technology that reduces ·pacing from a gun in the service of union-busting company agents: both investments ultirnatelv have the same effect and the economic results look the same on the balance sheet: As Jeremy Brecher reminds us. "The critical historian must go behind the economic category of cost-minimization to discover the social relations that it em bod ks (and conceals)" ( 1978). 1 ·· -~, J . ,.:,,, 115 Ever since the nineteenth century, labor-intensive machine shops have been a bastion of skilled labor and the locus of considerable shop-floor struggle. Frederick Taylor introduced his system of scientific management in part to try to put a stop to what he called "systematic soldiering" (now called "pacing''). Workers practiced pacing for many reasons: to keep some time for r.hemselves. to exercise authority over their own work. to avoid killing "gravy" piece-rate jobs by overproducing and risking a rate cut, to stretch out available work for fear of layoffs. to exercise their creativity and ingenuity in order to "make out" on "stinkers·• (pooriy rated jobs), and. of course. to express hostilit~· to management (see articles by Roy; Mathewson 1969). Aside from collective cooperation and labor-prescribed norms of behavior, the chief vehicle available to machinists for achieving shop-floor control over production was their control O\'er the machines. Machining is not a handicraft skill but a machinebased skill; the possession of this skill. together with control O\'er the speeds, feeds. and motions of the machines, enables machinists alone to produce finished parts to tolerance (Momgomery 1976b). But the very same skills and shop-floor control that made production possible also make pacing possible. Taylor therefore tried to eliminate soldiering by changing the process of production itself. transfer:-ing skills from, the hands of machinists to the handbooks of management; this. he thought, would enable management. not labor. to prescribe the details of production tasks. He was not altogether successful. For one thing. there is still no absolute science of metalcutting and methods engineers, time-study people. and Method Time '.\lcasurement (MTM) specialists-however much they may han: changed the formal processes of machine-shop practice-have not succeeded. in. puttiPg a stop to shop-floor control over production.* Thus, when sociologist Donald Roy went to work in a machine shop in the 1940s. he found pacing ali\'e and well. He recounts an incident that demonstrates how traditional patterns of authority rather than scientific management still reigned supreme: • The setting of rates on jobs in machine shops is still more of a guess than a scientific determination. This fact is not lost on machinists. as their tYpical descriprions of the methods-men suggests: "They ask their wh·es. they don't know: thev ask their children. thev don't know: so they ask their friends." Of courst", this apparent and ackno,..i°edged lack of scientific certainty comes into play during bargaining sessions over rates. when "fairness" and power. not science. determine the outcome. 1 _, , .. l.J 116 "I want 25 or 30 of those by 11 o'clock," Steve the superintendent said sharply. a couple of minutes after the 7: 15 whistle blew. I [Roy] smiled at him agreeably. "I mean it," said Steve, half smiling himself. as McCann and Smith, who were standing near us, laughed aloud. Steve had to grin in spite of himself and walked away. "What he wants and wha•. he is going to get are two different things," said McCann: (1953:513) . Thirty years later, sociologist Michael Bur;iwoy returned to the same_ shop_and concluded, in his own study of shop-floor relations, that "in a machine shop, the nature of the relationship of workers to their machines rules out coercion as a means of extracting surplus" ( 1976). This was the larger contex~ in which the automation of ma• chine tools took place; it should be seen. therefore, as a further managerial attempt to wrest control over production from the shop-floor work force. As Peter Drucker once observed, "What is today called automation is conceptually a logical extension of Taylor's scientific management" ( 1967:26). Thus it is not sur• prising that when Parsons began to develop his N/C "Cardomatic" system, he took care not to tell the union (the UAW) in his shop in Tra\"erse City about his exciting new venture. At GE (Schenectady). a decade of work-stoppages over layoffs. rate cuts, speed-ups, and the replacement of machinists with less s"killed apprentices and women during the war, culminated in 1946 in the biggest strike in the company's history, led by machinists in the United Electrical Workers (VE) and bitterly opposed by the GE Engineers' Association. GE's machine-tool automation project, launched by these engineers soon afterward, was secret, and although the r ·oject had strong manage· ment support, publicist Vonnegut rL~alled, with characteristic understatement, that "they wanted no publicity this time."* During the first decade of machine-tool automation develop· ment. the aircraft industry-the major user of automatic machine tools-also experienced serious labor trouble as the machinists and auto workers competed to organize the plants. The postwar depression had created discontent among workers faced with layoffs, company claims of inability to pay, and mas• sive downward reclassifications (Allen and Schneider 1956). Major strikes took place at Boeing, Bell Aircraft (Parsons' prime con• tractor), McDonnell Douglas. Wright Aeronautical. GE (Evan• dale) (jet engines), North American Aviation, and Republic Air· • Kurt Vonnegut . lt>tter to author. February 1977. 1 ., ff l.'i 117 craft. It is not difficult, then, to explain-the popularity among management and techni~al men of a November 1946 Fortune article entitled "Machines Without Men." Surveying the technological fruits of the war (sensing and measuring devices, servomechanisms, computers, etc.), two Canadian physicists promised that "these devices are not subject to any human limitations. They do not mind working around the clock. They never feel hunger or fatigue. They are always satisfied with .working conditions, and never demand· higher wages based on the ~ompany's ability to pay." In short. "they cause much less trouble tl.~n humans doing comparable work"' (Leaver and Brown 1946:203). One of the people who was inspired by this article was Lowell Holmes, the young electrical engineer who directed the GE automation project. However, in record-playback, he developed a system for replacing machinists that ultimately retained machinist and shop-floor control over production because of the method of tape preparation.* This "defect" was recognized immediately by ~hose who attended the demonstration of the system; they showed little interest in the technology. "Give us something that will do what we say. not what we do," one of them said. The defects of record-playback were conceptual, not technical; the system simply did not meet the needs of the larger firms for managerial control over production. N/C did. "Managers like N/C because it means they can sit in their offices, write down what they want, and give it to someone and say, 'do it,'" the chief GE consulti,"l~ engineer on both the record-playback and N/C projects explarned. "With N/Gthere is no need to get your hands dirty, or argue" (per~.onal interview). Another consulting engineer, head of the Industrial Applications Group which served as intermediary between the re5earch departn:ient and sales department at GE (Schenectady) and a key figure in the development of both technologies, explained the shift from record- • The fact that record-playback lends itself to shop-floor control of production more readily than N/C is borne out hy a study of N/C in the United Kingdom done hy Erik Christiansen in 1968. Only in thoS<' cases where record-playback or plughoard controls were in use (he found six British-made record-playback jig borers) did the machini\t keep the same pay scale as with conventional equip• rnent and retain control over the entire machining process. In Christiansen's words, record-playback (and plogboard programming) "mean that the shop flllor retains control of the work cycle th rough the skill of the man who first programmed the machine" (1968:27. 31). 115 118 pla>·back to N,C: "Look, with record-playback the control of the machine remains wich the machinist-control of feeds, speeds, number of cuts, output; with N/C thl'!re is a sr:{t of control to management. Management is no longer dependent upon the operator and can thus optimize the use of their machines. With NIC, control over the process is placed firmly in the hands of management-and \\-hy shouldn't we have it?" (personal interview). It is no wonder that at GE. NIC was often referred to as a management system, not a!> a technology of cutting metals.• Numerical control dovetailed nicely with larger efforts to computerize company operations, which also encailed concentrating the intelligence of manufacturing in a centralized office. In the intensely anti-Communist I 950s. moreover. ac: one former machine~tool design engineer has suggested, NIC looked like a solution co securi~y problems, enabling management to · remove blueprints frc11 che floor so thac subversives and spies couldn't gee their hands on them. NIC also appeared to minimize the need for coscly cooling and it made possible the culling of complex shapes that defied manual and tracer mechods. and reduced actual .:hip-cutting time. Equally important. however, ~IC replaced problematic rime-study methods wich "cape time"-using the time ic takes to run a cycle ~s the base for calculacing rates-replaced troublesome skilled machinists with more lractable "hutton-pushcrs." and eliminaccd once and for all the problem of pacing. If. with hindsight. ~IC seen•s to have led co organizational changes in the factory. changes which enhanced managerial control over production. it is because the te,hnology was chosen. in pan. for just that purpose. This becomes even clearer when we look al how the chosen technology was deployed. Choice in Deployment: Managerial Intentions There is no question bur thac management saw in NIC the potential to enhance cheir authoricy over production and seized upon it. despicc questionable cost effectiveness. t Machine-tool • GE Company 1958. See also Forrester ,1 al. 1955. t The cost effectiveness of N/C depends upon many factors. including training costs. programming costs, computer t:osts. and the like, beyond mere time saved in actual chip-cutting or reduction in direct labor costs. The MIT staff who 1-:, ,..., .1. tJ 119 conducted the early studies on the economirs of ~/C focused 011 the savings in culling time and waxed eloquent about the new revolution. At the same 1imC', however, ther warned 1ha1 the key 10 the economic vi,1hili1v of -:,.:.1c was a reduction in programming (softl,·are) <:os1s. Machine-10<>1 company salesnll'n were not disposed lO emphasize these po1emial drawbacks. though. and numerous users went bankrupt because they believed what they were told. In thl· early days. however, most users were buffered against such tragedy by s1a1e subsi,h. Today. potential users are some,~·ha1 more cautious. and machine-tool buildt·Vi are more restrained in their advertising, tempering thC'ir promise of economic success with qualifiers about proper use, the righ1 101 and batrh si1.C', suffi<:ie111 training, etc. For the independe111 imes1iga1or. it is extremely difficult 10 ;1,sess the economic viabili1v of such a technology. There are many rea~om for 1hi,. Firsr. the data is rarely available or ac-cessible. Whatever the mo1iva1ion--1ec:hnical fascination. keeping up with competitors. etc.-1he pur<:hase of new capital e4uipme11t must be justified in eco·nomic terms. But justifications are not 100 diffin1h tn come bv if the item is desired enough bv tht· right p,·nplc. They_ arc Sl·lfimerestcd anticipations and 1hu, u~ually op1imis1ic ones. ~!me important. firms rarely conduct pos1audi1s on their purchases. 10 see if their justific.:,11iom ,,·erT warra111ed. :--:ohody wants lO documc111 his error, am! if the machinen is tixecl in its foundation. that is whe1 l' ii will stay. whatever a pos1audi1 ren·als; ~ou learn I'> live with it. The poi111 here 1s 1ha1 the economics of c-apit;il l'111ed amnn~ dep.1r1111em,, with separate budget,. and the rn,t, to 11rn: ;in• 1:1e hidden costs Ill the others. Also. there is C'very r l•a~on lO bclil'nc" 1ha1 the data 1ha1 does exist is ~elf-serving information provickd b, each operating unit 10 enhance its position in the firm. And. fmallv. there i, the 1rirk\· 9ue~1ion nf h•>w "vi,1bili1v"' is defined in the fir,1 pla<:e. Sometimes. marhine, make moue~ for ,t <:ompanv 1,·he1her the, were u~ed produc1ivclv or not. The purpo,;t" of this asidt· is 10 emph.1si1e 1ht· fJn 1ha1 "houorn-line" expl;1- nations for complex hrs10ric;1l ckvdoprnc::111~. like the i111rnduc tion of 11t•\,· <:.1pttJI e9uipme111. are ne,·cr in 1hemseh·es suffi<:ic111, nor neces~anl\· 10 he 1rus1cd. If .1 com pan~· wams 1ojmmduce somC'thing new. i1 must jus1ifv it in lt·rm,; of mak111g a profit. This is not 10 sav. however. 1ha1 profit making wa~ its real (or. if ,;o, it~ only) motive or that a profit was ever made. In the case of au1oma1ion. ,1t·p, are taken le~~ nut of careful cakulati()n than <)n the faith 1ha1 it is alwav,; ~o.,d 10 replace capital with labor, a faith kindled deep in the soul of manufanuring engineer~ and managers (as economist Mi<:hael Piore. among others. has shm,·n. See, for example, Piore 1968). Thus, au1omat1on is dri1·e11 forward. not simply bv the profit motive. but by the ideology of au1oma1ion itself, whicl, 1 eflt·c1~ the· S the nt·w conditions: no foreme n or punch clock, their own tool crib. their own sd1eduling of parts through the sho p, and even some training in programming. ~torale improved and turno\·er. absenteeism, and the scrap-rate declined atcordinl<(ly. However. managerial enthusiasm for the experiment soon waned, and, after onlv a few half-hearted years. it was unilaterally called off. The company claimed that the union·s desire to extend thf" experiment to ocher areas of the shop and to othl plants within the same corporation threatened to make llu 122 A second illustration of the managerial imperative behind technological determinism is provided in an intervie\\' I had with two shop managers in a plant in Connecticut. Here, as elsewhere. much of the N/C programming is relatively simple, and I asked the men why the operators couldn't do their own programming. At first they dismissed the suggestion as ridiculous. arguing that the operators would have to know how to set feeds and speeds. that is. be industrial engineers. I pointed out that the same people probably set the feeds and speeds on conventional machinery. routinely making adjustments on the process sheet provided by the rrtethods engineers in order to make out. They nodded. They then said that the operators couldn't understand the programming language. This time I pointed out that the operators could often be seen reading the mylar tape-twice-removed information describing the machini;,z being done-in order to know \•,hat was coming (for instance. to anticipate programming errors that could mess things up). Again. they noddt>d. Finally they looked at each other. smiled. and cne of thein leaned over and confided. "We don't want them w:· Here is the reality behind technological determinism in deployment. Reality on the Shop Floor Although the evolution of a technology follo\,·s from the social choices that inform it. choices which mirror the social relations of production. it would be an error to assume that in having exposed the choices. we can simply deduce the rest of realitv from them. Reality cannot be extrapolated from the intentions that underlie the technology any more than from the technology itself.* Desire is not identical to satisfaction. "In the conflict between the employer and employed.'" John G. Brooks obsened in 190:,, "the 'storm centre· is largely at this point where science ind invention are applied to industry."t It is the program 100 expensi\'e since an extension of the experiment meant also an extension of the bonus. The union business agent. former!,· a shop steward in the experimental program and one of it~ s1auchest supporters. l"Xplained the termination in another wav: the companv was losing control over the wt>rk force. • This is an error that Bra\'erman tended to make in discussing NiC. t Cited in D. ~lontgomen (unpublished: Ch. 4. p. I}. 123 here that the reality of N/C was hammered out, where those \,·ho chose the technology finally came face-to-face with those who did not. The introduction of N/C was not uneventful, esr~cially in plants where the machinists' unions had a long history. Work stoppages and strikes over rates for the new machines were common in the l 960s, as they still are today. At GE, for example, there were strikes at several large plants and the entire Lynn, Massachusetts plant was shut down for a month dm·ing the winter of 1965. There are also less overt indications that management dreams of automatic machiner~· and a docile, disciplined work force but they have tended to remain just that.t t Perhaps the single most important , and difficult, task cnnfron_!ing the critical student of such rapid(~ evolving 1echnologin as ~/C is 10 try to dist·manglc dreams from realities. a hopc:·, who refer to the automatic lnadin~ of :--;tc m:Khincs by th<: l'nim:,te robots. to Flexible ~lanufacturin~ S,·stems (F~IS) that tic any num()(."r of machines together with an automatic transfer line, to adaplivt· corllwl, ,virh sensors that automatically correct for IOol wear and rough c:is1i11gs and !he like. or to Dirc-t·t ~umerical Control sntcrns (D!\C) which cc:ntrali1c corllrol u,-cr a whole pl,1111 of ;lj/C equiprnelll through one computer. Three important thin~, must be kept in mind 1,·ht·n dc-aling with such rou111erargume11t~. First. terhnical people. i1 mu~r be- rc-1:H·tnbaed. alwavs havc tht·ir eyes on tht' future-it is tht·ir job; the~ live in the statt·-of-the-Jn world which of1,·n has n-rv little conneni<>n with indusrrial realit,·. Thus. it rs hardh· rnrprising that 1erhnital forecasters of the late I 95(h prt•dicted that bv now JI le,1~1 i:, pc-rn·111 of machinc tools in thisn,untn· 1,ould be ~/C(it is less than'.? pt·rre111). and that we would tic seeing fullv automatic metalworking fartories (r here an· nont·). There is no better reason 10 belic-ve tht" enb•incering and track journals tocia) . much less the self-serving forerasts of manufacturing engirwers. All too often. social anah·,1, mereh- echo these prophets. extrapolating wondt·rful or wcwful n,nset1uencn of projected technologiral chanv;t's without pa\"ing the slightest arwntinn to the mundane ,icissitudes 0f histllriral cxpe1·iem·c. or industrial practice. To them. the critic must respond: look again . Second. judging from past experience. there i, littlt- rea~on ~implv 10 ;1s~umt· that the new experimental or demonstration sv~tcm, will .1ctu.11ly funnion 011 rlw shop floor as intended, mu(·h ll'ss perform economically. fhis author has visited 121 124 Here we will examine briefly three of management's expectations: the use of "tape time" to set rates; the deskilling of machine operators; and the elimination of pacing. Early dreams of using tape time to set base rates and measure performance and output proved fanciful. As one N/C operator observed, while rates on manual machines were sometim-:s too high, they were usually within a reasonable range, whereas the rates on N/C wer,e "out of all relation to reality-ridiculously high; N/C's were supposed to be like magic but all you can do automatically on them is produce scrap." The machines, contrary to their advertisements, could not be used to produce parts four plants in the United States with FMS systems and found their economic justifications suspect, their down time excessive, and their reliability heavily dependent upon a highly skilled force of computer operators, system attendants, and maintenance men; there was also little sign of further development. Adap- .tive systems, under development at Cincinnati Milacron, are still in an experimental stage; when placed on the shop floor, these even more complex and sensitive pieces of machinery are bound to produce more maintenance problems than they solve. DNC is simply another name for the automatic factory, the supreme fantasy of the industrial technocrats, now heralded by self:serving computer jocks, supported by beleaguered corporate managers (whoSc farsightedness is more rhetorical than real), and, as usual. funded by the military (in this case. the air force ICAM program). Third, the ultimate viability of these technologies under the present mode of production depends, in the final analysis, upon the political and economic conditions that prevail and upon the relative strengths of the classes in their Slruggle over the control of production. To assume simply that the future will be what the designers and/or promoters of these technologies think it will be. would be to beg all of the questions being raised here, to ratify. out-of-hand, a form of technological determinism. Further, it would be to deny the realm of freedom that is being described, a freedom which could result not only in the delaying or subverting of these technologies (and thus the purpos~s they embody}-allowing for more time to struggle for greater freedom-but also in the fundamental reshaping of their design and m: to meet ends other than simple capital accumulation and the extt:nsinn of managerial and corporate power. See, for example. the discussion of Computer Numerical Control (CNC) in the final section on "alternative realities." In short, a facile reference to the future is the educated habit of technical people in our society. people who are quite often seriously (and sometimes dangerouslv) ignoram of the past and mistaken about the present. Tn ado?>t their habit would be to suspend _judgment (or, r.:,ther, yield to their judgment), to forego the critical, concrete, historical examination and assessment of the presem situation, which alone can guide us intelligently into the still clouded future. 122 125 to tolerance without the repeated manual inter:,.vention of the operator in order to make tool offset adjustments, correct for tool wear and rough castings, and correct programming errors (not to mention machine malfunctions, such as "random holes" in drills and "plunges" in milling machines, often attributable to overheating). As the N/C operator just quoted explained, in a response to a New York Times article on the wonders of computer-based metalworking: Cutting metals to critical tolerances means maintaining constant control of a continually changing set of stubborn, elusive details. Drills run. End mills walk. Machines creep. Seemingly rigid metal castings become elastic when clamped to be cut, and spring back when released so that a flat cut becomes curved, and holes bored precisely on loc.:tion move. somewhere else. Tungsten carbide cut• ters imperceptibly wear down, making the size of a critica,l slot half a thousandth too small. Any change in any one of many variables can turn the perfect part you're making into a candidate for a modern sculpture garden, in seconds. Out of generations of dealing with the persistent, ornery problems of metal cutting comes the First law of Machining: "Don't mess with success." (Tulin 19i8:16) In reality, N/C machines do not run by themselves-as the Un:ted Electrical Workers argued in its 1960 Guidi' to Autonwtion, the new equipment, like the old, requires a spectrum of manual intenention and care[ul attention to detail, depending upon the machine, the product, and so on. The fiction that the time neces::-ary to do a job could be determined by simply adding a standard factor or two (for setup. breaks. etc.) to the tape (cyc:le) time, was exploded early on, and with it hope of using the tape to measure performance (although some methods people still try). The deskilling of machine operators has also, on the whole , not taken place as expected, for two reasons. First, as mentioned earlier, the assigning of labor grades and thus rates to the new machinery was, and is, a hotly contested and unresolved issue in union shops. Second, in union and nonunion shops alike. the determination of skill requirements for N/C must take into account the actual degree of automation and reliability of the machinery. Management has thus had to have people on the ma<:hines who know what they are doing simply because the machines (and programming) are not totally reliable; they do not 1 ,., ("", I., tJ 126 L, run by them~dves and produce good finished parts. Also, the machinery is still very expensive (even without microprocessors) and thus so is a machine smash-up. Hence, while it is true that many manufacturers initially tried to put unskilled people on the new equipment, they rather quickly saw their error and upgraded the classification. (In some places the most skilled people were put on the N/C machines and given a premium but the lower formal classifications were retained, presumably in the hope that s9meday the skill requirements would actually drop to match the classification-and the union would be decertified.} The poinc is that th~ intelligence of production has neither been built entirely into the machinery nor been taken off the shop floor. It remains in the possession of the work force.* This brings us, once again. to the question of shop-floor control. In theory, the programmer prepares the tape (and thus sets feeds and speeds, thereby determining the rate of production), proofs it out on the machine, and then turns the show over to the operator,"who from then on simply presses start and stop buttons and loads and unloads the machine (using standard fixtures). This rarely happens in reality. as was pointed out above. Machining to tolerances generally requires close attention * The shortage of skilled manpo\.,er has always been cited by managers and technical people as a justification for the introduction of labor-saving technologies like :--1/C. Rarely, however, is the shortage actually demoi1strated or explained in any compelling way; it remains a necessary and unquestioned ideological prop. for a manpower shortage is a relative thing; relative to new air force and aircraft industry requirements in the cold war. there was a perceived shortage. But. given that shortages are only perceived in relation to a present or future need, they are predictable; they are not natural phenomena but socially created ones, remediable through training programs and sufficient monetary and other incentives. (This author remembers, for example, that not so long ago he went to college on loan programs created to deal with a recognized shortage of college teachers, relative to a vastly expanding educational system.) Thus, when managers introduce N/C because of the impending retirement of the last generation of skilled machinists, we must ask, where arc their replacements? Why have apprcntireship programs been eliminated or shortened? Why do vocational courses habituate young people to "semiskilled" work in the name of training for a craft:- The answer is that the shortage is. in reality, crea,ed to complement the new technology, not the other way around. Fortunately for capital. ho"·ever. the skill is not entirely eliminated. however "unskilled" the classification; passed on informally and on the job, it remains on the shop floor. If it wasn't there, finished parts would never make it out the door. 124 127 to the details of :.he operation and frequent manual interventiot'l through manual feed and speed overrides. This aspect of the technology, of course, reintroduces the control problem for management.Just as in the conventional shop, where operators are able to modify the settings specified on the worksheet (prepared by the methods engineer) in order to restrict output or otherwise "make out" (by running the machine harder), so in the N/C shop the operators are able to adjust feeds and speeds for similar purposes. Thus, if you walk into a shop you will often find feed-rate override dials set uniformly at, say. iO or 80 percent of tapedetermined feed rate. In · some places this is called the "iO percent syndrome"; everywhere it is known as pacing. To combat it, management sometimes programs the machines at 130 percent, and sometimes actually locks the overrides altogether to keep the operators out of the "planning process." This in turn gets management into serious trouble since the interventions are required to get the parts out the front door. It is difficult to assess to what extent the considerable amount of intervention is attributable to the inherent unreliabilitv of the · complex equipment itself, but it is certainly true that the technology de\'el<>ps shortcomings once it is placed on the shop floor, whether or not they were there· in the original designs. Machines often do not do what they are supposed to do and down time is still excessive. Technical defects, human errors. and negligence a1 e ackno\\'ledged problems. and so is sabotage. "I don·t care how many computers you have, they'll still have a thousand ways to beat you," lamented one manager of N/C equipment in a Connecticut plant. "When you put a guy on an N/C machine, he gets temperamental." another manager in Rhode Island complained. "And then, through a process of osmosis, the machine gets temperamental." On the shop floor, it is not only the choices of management that have an effect. The same antagonistic social relations that. in their reflection in the minn,; of designers. gave issue to' the new technology. now subvert it This contradiction of capitalist production presents itself to 1.1anagement as a problem of "worker motivation," and management's acceptance of the challenge is its own tacit acknowledgment that it de>es not have shop-Hoor control over production, that it is still dependent upon the work force to turn a profit. 1 () - .(,t_} 128 Thus, in e••'.lluating the work of those whose intentions to wrest control over production from the work force informed the design and deployment of N/C, we must take into account an article written by two industrial engineers in 1971 entitled "A Case for Wage Incentives in the N.C. Age." It makes it quite clear that the contradiction of capitalist production has not been eclipsed--computers or no computers: Under automation, it is argued, the machine basically controls the manufacturing cycle. and therefore the worker's role diminishes in importance. The fallacy in this reasoning is that if the operator malingers or fails to service the machine for a variety of reasons, both utilization and subsequent return on investment suffer dras• tically. . Basic premises underlying the design and development of N.C. machines aim at providing the capability of machining configurations beyond the scope of conventional machines. Additionally. they "'de-skill"' the operator. Surprisingly, however, the human element continues to be a major factor in the realization of op- timum utilization or yield of these machines. This poses a continuing problem for management. because a maximum level rf utilization is necessary to assure a satisfactory return logv lends 'Y./C as never before to total shop-floor control. Although th<: large metalworking plants in the United States are steadik introducing C~C equipment, the potential for shop-Honr comrnl is far from being realized. The GE plant in Lynn, ~Ltss;1chusetts. is a typical example. Here machine operators art.· W>l permitted to edit programs-much less to make their own--011 the new CNC machines; quite often the controls arc locked. Only supervisory staff and programmers are allowed to edit the programs. Managers are afraid of losing shop-floor rnntrnl or confusing their tidy labor classification and wage svstem: prngrammers are afraid that operators lack the 131 120 training and exr~rience required for programming-an argument that has convinced at least so11e operators that these functions are beyond their intellectual grasp. The shortcomings of this system for the operators are obvious. Less obvious are the shortcomings for management: lower quality production and excessive machine down time. If the programs are faulty and the operator cannot (or is not allowed to) make the necessary adjustments, the parts pt"oduced will be faulty . If a machine goes down because of programming problems on the second and third shifts, when the programmers are not around, it is.likely to be down for the night, with a corresponding loss in productivity. The situation is quite different in the state-owned \\·eapom facto_ry in Kongsberg, Norway. a plant with rouehly the same number of employees, a similar line of products (aircraft parts and turbines). a similar mix of commercial and military customers, and, most important, the same types of Ct,;C machinery (although here they tend to be European-made rather than Japanese) as at GE.* But in Norway the operators routinely do all of the editing, according to their own criteria of safety, efficiency. quality, and convenience; they change the sequence of operations, add or suhtract operations, and sometimes alter the entire structure of the program to suit themselves. \Vhc:n they are satisfied with a program and have finished producing a batch of parts, they press a button to generate a corrected tape which. after being approved by a prograrnrner, is put into the librarv for permanent storage. All operators are trained in ":-1/C programming and. as a consequence, their conflicts with the programmers are reduced. One programmer-who, like most of his colleagues, had received his training in programming while still a machine operator-justified having any programmers at all by the fact that the programmer was a specialist and was thus more: proficient (he also dealt directly with customers and did most of the APT programming of highly complex aircraft parts). Yet when asked if it bothered him to have his welJ-worked programs tampered with by the operators, he replied. without hesitation. that "the operator knows best: he's the one who has to actually • Tht· follm\'ing discussion of the situ,11ion i11 Kongsherg, :-.orway, is hased upon corrcspondcm:c and personal contact with par1icip;,i11s in the trade union participation project and a recent research visit to Scandinavia (October 1978). 123 132 make the part :ind is more intimately familiar with the particular safety and convenience factors; also, he usually best knows how to optimize the program for his machine." This situation, it should be F-" )inted out, is unusual even for Norway. It is the result of many factors. The Iron and Metalworkers' Union in Non\'ay is the most powerful industrial union in the country and the local "club" in Kongsberg is a potent force in the industrial, political, and social life of Kon$sberg, representing a cohesive and rather homogeneous workmg-cfass community. The factory is important in state policy, as a holding company in electronics, and is an important center of high technology engineering. Also, social democratic legislation in Norway has encouraged worker participation in matters pertaining to working conditions and has given unions the right to information. Most important, however, the local "club" has been involved for the last seven years in what has been called the "trade union participation project," an important development in workers' control which focuses upon the introduction of computer-based manufacturing technology. In I 971, the Iron and Metalworkers' Union, faced with an unprecedented challenge of new computer-based information and control systems (for production, scheduling, inventory. etc., as well as machining), took steps to learn how to meet it. They succeeded in hiring. on a single-party basis (that is, without management collaboration). the government-run Norwegian Computing Center to research the new technology for them. As the direct result of this unprecedented effort, computer technology was demystified for the union, and the union-and labor in general-was demystified for the computer scientists at the Center; the union became more sophisticated about the technology and the technical people became more attuned to the needs and disciplines of trade unionists. In practical terms, the studv resulted in the production of a number of textbooks on the ~ew technology, w·ritten by and for shop stewards. the creatir,n of a new union position. the "data shop steward," and, in time, the establishment of formal "data agreements·• (between indi\'idual companies and their local "clubs" and between the natioral union and the employers' federation) which outlined the union's right to participate in decisions about technology. The Kongsberg plant was the first site of such trade union participation. Herc the data shop steward. a former assembly worker. is responsible for keeping abreast of and critically 133 1:; 0 scrutinizing al! new systems; another man is assigned the job of supervising the activity of the data shop steward to ensure that he doesn't become a "technical man," that is, captive either of the technology or of management and out of touch with the interests of the people on the shop floor. The responsibilities are enormous: this is not a situation in which union and management cooperate harmoniously, nor is it a management-devised job-enlargement scheme to motivate workers. The task of the data shop steward. and the union in general. is to engage. as effectively as possible, in a struggle over information and control. a struggle engaged in, with equal sophistication and earnestness, by the other side. \\'hen manageinent plans to introduce a new computer-based production system, for example, the union must assume as a matter of course (based upon long experience) that the proposed qesign reAects purposes that are not necessarily consonant with the interests of the workers. The data shop steward and his colleagues must learn about the system early enough. and investigate it thoroughly enough, to ensure that it contains no features that make possible. for example, the measurement of individual performance or any monitoring of shop-floor activities that would restrict worker freedom or control. As it turns· out. all new svstems invariablv contain such features (since thev are often ca,;10uflagcsibilities remain.