Automation, Skill and the Future of Author(s): Paul S. Adler Reviewed work(s): Source: Berkeley Journal of Sociology, Vol. 33 (1988), pp. 1-36 Published by: Regents of the University of California Stable URL: http://www.jstor.org/stable/41035388 . Accessed: 25/05/2012 16:19

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Regents of the University of California is collaborating with JSTOR to digitize, preserve and extend access to Berkeley Journal of Sociology.

http://www.jstor.org ,Skill and the Future of Capitalism

By Paul S. Adler

1. Introduction

The area of "labor process studies"has undergoneimportant mutationssince Braverman's (1974) landmarkstudy Labor and Monopoly Capital Braverman's basic thesis was that capitalist domination of the labor process has been progressively consolidatedby a deskillingand degradationof work and thatthis same transformationof work would ultimatelylead to revolt by alienated workers.Against Braverman's diagnosis of "a secular trendtoward the incessant lowering of the workingclass as a whole below its previous conditionsof skill and labor" (1974, p. 129), the more recentinclusion of a varietyof other factors- market pressures, ideologies, institutions,worker resistance-as counterweightsto deskillingpressures helps explain the greater diversityof outcomesthan Braverman's approach would imply.

The resultantstate of research-as exemplifiedin collections like The Degradation of Work? (Wood, 1982), Job Redesign: Critical Perspectiveson the Labour Process (Knights,Willmot, and Collinson, 1985), or in the research of Noble (1984) and Shaiken (1985)-has gained enormously in resolution and precision.

But, in the process,a key challenginginsight of Braverman's work has been abandoned.The idea thatclose studyof the labor process could yield insight into the destiny of capitalism is increasinglyconstrued as futile.Whereas Braverman saw trendsin work contentand organizationas shaping the long-runevolution of capitalism-defined very broadly as a formof societybased on the competitionof independentfirms employing wage-labor- the more recent research aims at finer-grainedanalysis of specific firms or industries.It is thereforedeliberately modest in its historicalscope.

These new questionsare indeed veryinteresting; but I do not believe that they exhaust the issue. Moreover,I shall attemptto show thattheir analysis can benefitfrom the restorationof more viable generalizationsat the broader,historical level.

This article goes thereforesomewhat against the current. Many commentatorshave argued that it is time to leave Bravermanbehind By contrast,my critiqueof the shortcomings of Braverman's analysis of long-run,aggregate tendencies will 2 BERKELEY JOURNAL OF SOCIOLOGY retainhis ambitionof historicalscope, and will attemptto build a conceptualbasis for useful empiricalresearch that tears directly on the "big questions"he addresses.

To summarizethe key pointsof my argument:the difficulty with the automation-skilldebate as it pertainsto long-termtrends lies in the concepts of skill being used. Braverman's skill yardstick~thecraftsman-is increasingly obsolete (Section 2). An alternativeframework for analyzingskill is developed fromclose examinationof the case of bank computerization(Sections 3 and 4). This frameworkencompasses a quantitative dimension- substantivecomplexity as measuredby training-timerequirements- -and three qualitativedimensions- responsibility, abstractness and interdependence(Section 5). Compared to Braverman'sfocus on craft autonomy and expertise,this four-dimensionalframework provides a richer frameworkfor empirical investigationof skill trends.These dimensionsalso prove to be homologous with the basic buildingblock of the capitalistform of society-thatis, the commodity,the productproduced for sale on the market(Section 6). This frameworkthus allows us to addressdirectly the question of how skill trendsmight impinge on capitalism'sviability.

The bankingcase (and manyother case studiesre-read in this light) suggeststhat competitive pressures force managersto seek out more productiveways of implementingautomation, and these ways typically(although not always) involve more training,higher levels of responsibility,more abstracttasks and goals, and greater functionalinterdependence. My argumentwill be thatthese trends in automation's impact on skill requirementsundermine the effectivenessof traditionalpersonnel management practices. In doing so, these trendssimultaneously undermine the viabilityof the commodityform of labor-powerand create the premise of alternativepost-capitalist forms of organization.

2. The Need for a New Concept of Skill

The centralproblem is that of the impact of automationon skill requirementsunder capitalist conditions. My focus is exclusively on long-run and aggregate trends.1This section explainswhy we need a new conceptof skill.

1. Thereare, of course,many crucial issues to be exploredin the shortrun: in particular,under capitalist conditions, structural shifts are generallynot plannedat the societallevel, with the effectthat even if theseshifts are positivefor most workers, and a fortioriif theshifts are only. favorablein the long run,they impose real costs on individual workersand wastesocial resources. ADLER: AUTOMATION AND CAPITALISM 3

Spenner's (1983) review of the major empirical studies on skill trends is most useful. The result of this survey is a resoundingrejection of Braverman'sdiagnosis of "a secular trend toward the incessantlowering of the workingclass as a whole below its previousconditions of skill and labor." Not one of the systematic,aggregate studies of trendsin individualoccupations or in average labor-forcerequirements shows a deskillingtrend; most show a clear up-grading both for most (although not all) occupationstaken individually and forthe labor forceas a whole.

But the slipperycore of all these studies~and the core that Bravermanpre-emptively attacked in the concludingchapter of his book-is in our measurementof skill. The basic reason for this slipperinesslies not so much in the tools of observation(although those problemsare substantial),as in our concepts of skill. The literatureis populatedby almost as many skill concepts as there are authors(see referencesin Spenner,1983 fora briefoverview).

Braverman'sconcept of skill has indubitableattraction. It is based on a craft model, encompassingtask variety and scope, responsibilityand autonomy,and the integrationof mental,manual and social componentsof work. It can be summarized as a conceptwith two key dimensions:substantive task complexityand autonomy/control;but in Braverman's view these are closely correlated.Indeed such a close correlationis criticalto his whole argument.

That such a correlationis not self-evidentis, however,easily seen. One can, for example, easily imaginejobs with almost no autonomybut veryhigh trainingrequirements- and thatone would have to qualifyas highlyskilled: take the airlinepilot who must follow a pre-setprocedure for every situation.On the otherhand, one mightwonder what one should make of an infantrymanin the military,who is a jack-of-all-trades,but who is considered-for having full masteryof a complex, multi-facetedtask and a low degree of specialization-tobe the lowest-rankingelement in the skill hierarchyof the military.

It is thereforeimportant to distinguishmore carefullythe different"dimensions" of skill. If our focus is on the very long run,then it mightbe appropriateto begin our search for a viable skill concept with the economic determinantsof skill. In this perspective,skill is basically the relative economic value of differenttypes of labor; on this basis, skills can be measuredby the relativetraining-times associated with their"production." This dimensioncould be thoughtof as closely related to Braverman's substantivetask complexity dimension. By relating it to an economic mechanism,we can assure ourselves that our construct is not only a thought-construct,but also one that reflectsa real process-that of the basic, long-run "value" determinationof 4 BERKELEY JOURNAL OF SOCIOLOGY relativewages.2 It remains to be seen, of course, how relevant this determinationis, relativeto the weightof social and ideological forces acting on relativewages. There are excellent reasons for thinkingthat in anythingbut the long run supply-and-demand, institutional,political and ideological realities can swamp this economic determinationof skill. Moreover,which characteristics individual firms recognize as skill varies greatly-withsome constraintslike the relativeundervaluing of women's jobs. But in a theoryof long-runhistorical tendencies, it mightbe usefulto at least begin by abstractingfrom these qualifications.

Now, insofaras skill is basically a questionof trainingtime, the data reviewed by Spenner suggeststhat skill levels (as best they have been measured) have increased over time. It is, of course, impossibleto tell fromthe data whetheremployers have merely profited from exogenous trends in the workforceas regards,in particular,rising educationallevels, or whetherit is primarilythe requirementsof productionthat have drivenchanges in the trainingand educationalsystems. But in eithercase, over the long run, skill-understoodas trainingtime-and automation bear some positive correlationto each other. (Note that this propositionin no way excludes the possibilitythat workers' skills are risingeven fasterthan job skill requirements.)

Whateverthe weaknesses of this training-timedefinition of skill, it has one key strength,namely, that it poses with great claritythe next step in the analysis.If skill is definedin this way, it has a purely quantitativemeasure, being ideally reducible to hoursof trainingtime. The questionthat immediately makes itself manifestis then:what sortof qualities is the trainingdesigned to inculcateor to elicit? The challenge,therefore, is to establish a frameworkfor analyzing this "qualitative"question. This has been one of the attractionsof Braverman's framework:his yardstickfor skill-the craftsman-exemplifies not only the quantitativedimension I have identifiedas substantive task complexity,but also an intuitivelyappealing qualitative dimension-autonomy/control.

2. This premisereflects a certainconvergence of humancapital theoryand classical Marxisttheory. For Marx, the relativevalue of differentclasses of thecommodity "labor power" depends on thesocially necessarylabor time required to "produce"it, i.e., trainingand education. Its "price"will fluctuatearound this value as a functionof short-term supplyand demand factors, perhaps systematically biased by politicaland ideologicalfactors. The humancapital approach is thusa modelof the surfacerelations characterizing the marketfor labor power- a price model,as opposedto theunderlying value model. ADLER: AUTOMATION AND CAPITALISM 5

Braverman'scraft concept of controlis, however,inadequate. The key weakness of Braverman's craft yardstickis that the dissolution of craft autonomy is no longer central to the comprehensionof contemporarytrends. Ignoring the fact that a craft status never characterizedmore than a small minorityof workers(Form, 1980), we mightaccept thatthe fragmentationof craft was a central featureof the passage from handicraftto .Although different industries, regions and plants evolve at unevenrates, the dominantprocess of twentiethcentury capitalism is, however, a new one-the passage from manufacturingto large-scale machine-based industry and the growing importance of technical change as a source of productivityimprovement. In otherwords, the dissolutionof craft and the associated forms of worker autonomycharacterizes an "early"phase of capitalistdevelopment-one that is certainlystill underway in some industries,but no longer representativeof contemporarycapitalism's basic thrust.In Marx's terminology,the dissolutionof crafthas more to do with the transitionfrom the "formal"subordination of labor to the "real" subordinationof labor thanwith the subsequentdevelopment of the real subordinationof labor.

In anticipationof some of these difficulties,Braverman argued: I hope no one drawsfrom this conclusion that my views are shapedby a nostalgiafor an age thatcannot be recaptured. Rather,my views about work are governedby nostalgiafor an age thathas notyet come into being, in which,for the worker, thecraft satisfaction that arises from conscious and purposeful masteryof the labor process will be combinedwith the marvelsof scienceand theingenuity of engineering,an age in whicheveryone will be able to benefit,in somedegree, from thiscombination (1974, p.7). "Craftsatisfaction" may not, however, "combine" so easily with the marvels of science. Think of the modern doctor: the advances of medical science-and not merely the capitalist organization of medicine-have led to a fantastic degree of specialization and an almost total loss of autonomy. This specialization has not, however, meant a deskilling-on the contrary.It is true that no one doctor enjoys "conscious and purposefulmastery of the labor process." (Unless, that it, "labor process" is tautologicallyredefined to be whatevercluster of tasks the individualdoctor does have controlover.) But this reflectsan increasinglycomplex interdependenceof differentelements of what mightbe called the "collectivemedical worker."The idea of craftsman-like"autonomy" and "control"cannot capture the quality of highly-skilledbut highly-specializedjobs, whether they be neurosurgeonsor manufacturingtechnicians. 6 BERKELEY JOURNAL OF SOCIOLOGY

However, Braverman's approach has another attractionthat we cannot ignore. With his conceptionof skill, he manages to pose a questionthat I have put at the heartof this essay, namely the question of capitalism's viabilityas a system.If, througha process thatwas inevitablegiven capitalistrelations of production, workerswere to become progressivelydeskilled and theirjobs were to become progressivelydegraded, then the implication would be clear: revolt would become progressivelymore likely. This increasingprobability of revoltwould also be fueled by the fact that deskillingwould break down hierarchicallevels within the workingclass, and thus formthe homogenousproletariat that Marx envisaged.

Now, phrasedthis way, we see a very interestingfeature of Braverman'sframework, namely, that it is formallyanalogous to the now out-modedtheory of "absoluteimmiseration:" just like the supposed tendencytowards the loweringof real wages, deskilling progressivelydegrades the workers'condition to the point where they have no choice but revolt. This implicit message has, I believe, been at the heartof the attractionof Braverman'swork. It is a powerfulpolemical stance.3This explains why it has been sustainedfor so long by so manyin the face of so much data that apparentlycontradict it.

But the principlethat a fruitfultheory of skill mightderive from a conceptual structurethat linked work qualities to the questionof the viabilityof capitalismis most intriguing.

At this stage of our searchfor a frameworkfor the qualitative analysis of work useful for long-runanalysis-a frameworkthat mighthave a bearingon the viabilityissue itself-we can proceed eitherinductively or deductively.The followinganalysis proceeds inductively.It outlinesthe major loci of the qualitativechanges in workthat I have foundin Frenchbanks. Afterbriefly outlining the major technological changes in banking, I will discuss the evolution of skill requirements, using a "pre-theoretical" categorization of job characteristics.On this basis, a new frameworkemerges quite naturally.I shall returnto sketch the deductiveroute and the implicationsof this frameworkin later sections.

3. The polemicalpower of the idea of absoluteimmiseration is bolsteredby thefact that there are so many"local" instances of absolute immiserationboth in theeconomic sense (cyclical downturns) and in the labor processsense (real cases of deskilling.)This is, no doubt,why Marx maintainsthe polemiceven in the passagesof Capital wherehe distanceshimself from the theory (Marx, 1981, p.768ff.). ADLER: AUTOMATION AND CAPITALISM 7

3. Automationin Banking Let us firstreview the natureof bank automation.I shall take my examples from French banking, with which I am more familiar(Adler, 1981).4 Over the 1950-1980 period,the numberof bank accounts was multipliedby a factor of more than five, encouragingFrench banks-which are, on average, monstersby U.S. standards5--toundertake very ambitious computerization programs.The resultingsystems in Frenchbanks are quite similar to those found in larger and more technicallyprogressive U.S. institutions.

Banks began introducingcomputers during the 1950s, even while mechanizationwas still being completed.Data processing was initiallyin the "batch"mode, with overnightprocessing; the liaison between the computercenter and the user-departments involved card-punching(or, later, magnetic tape recording) for data-entryand the physicaltransportation of computerprintouts.

The second phase of computerizationwas characterizedby the diffusionwithin the bank of computerterminals. First, they were installedin the processingand accountingdepartments. Then, towards 1970, they appeared in purely interrogativemode (data could be accessed but not entered)in the branches,where they offered up-to-datebalance information.Beginning in the late 1970s, interactiveterminals allowing data-entryas well as data- access began to appear in all the bank departments-adevelopment thatwas almostcomplete in the major Frenchbanks by 1985.

With the new interactivetechnology came changes in work organization. Managementwas quick to seize upon the potential offeredby the technologyto attackthe sources of growinglabor tensionin the "paperprocessing" departments. These tensionswere manifestedin France by a major, cripplingstrike in 1974; in the U.S. the correspondingtensions took the form of exacerbated turnover rates.6 The specialization of data-entrytasks also

4. My field work was conductedprincipally in one of France's big fourbanks over the 1980-1981 period. This researchinvolved 4 months full-timefield work, several hundredinterviews of bank personneland managementat all levels, and participantobservation for one month workingas a tellerat two branches,representative of Stage 3 and Stage 4 configurations,discussed below. Sections 3 and 4 draw on my discussionin Adler (1985).

5. Four Frenchbanks long figuredin Dunn and Bradstreet'stop ten worldbanks (by balance sheet).

6. O'Brien (1968) cites annual turnoveramong bookkeepersin U.S. banks in the 40% to 80% range. 8 BERKELEY JOURNAL OF SOCIOLOGY generatedintolerably high error rates in bothcountries. Frenchbanks thereforeabandoned the doctrineof centralizing processing activities and separating them from commercial functions.The new principlewas to de-specializedata-entry as far as possible, makingeveryone responsible for the accuracyof their own data. Processingwas progressivelydecentralized to the back- officeof the local branchand even rightout to the frontdesk. In enteringdata, the teller,like the otherusers, is now automatically informedof any inconsistenciesor implausible data that the systemhas detected,and is asked to rectifythe entry.Processing timeand errorsare therebyminimized.

As banking moved from mechanizationto computerization, work has been totallytransformed. The workeris now entirely dependent on the computer system-whose malfunctioningis frequentenough to constitutein itself a permanentfeature of computerizedwork. Work is mediated by a new language of computercodes. A series of tasks formerlyconsidered the very essence of bank work has been eliminated,including accounting imputationand adjustment,classification of documents,multiple entriesof data, manual data search,and supervisionby signature. A new range of tasks has been introduced.Accountants now diagnose and rectifyanomalies listed by the computersystem. New types of errors~and fraud-appear.Their greatercost and more difficult diagnosis are congruent with the level of automation.Surveillance of work no longer takes the formof a document-by-documentverification and the signatureby officers of large-sumtransaction records. It has become a combinationof the computercontrol of data consistency/plausibilityand personal identificationcode status,and the supervisor'sen masse, ex post surveillanceof a computerrecord of exceptionaloperations.7

The parallels between the computerizedon-line banking systemand the automationof petrochemicalrefineries are striking. If in refineriesproduction is of the "continuousflow" process kind,in banks thereis a literallyinstantaneous production process. Once the raw material-data- enters the bank's recordsin any form at any place, it is automaticallyand instantaneouslyfed into all the pertinentaccounts includingthe client's account, the bank's generalledger, its auxiliaryaccounts, official accounts, and control

7. Access to the data systemis obtained by keying-ina personal identificationcode, which also indicatesto the computerthe natureof the operationsthe operatoris authorizedto conduct. It is the use of such codes that makes individualjob monitoringtechnically possible. Note thatin France the threatof labor strifeencouraged bank managementsto abandon any such monitoring. ADLER: AUTOMATION AND CAPITALISM 9 accounts.8

With this automation,banking finds itself confrontinga general characteristicof higher levels of automation.Reducing repeatedentry of the same data and controllingautomatically for data plausibilitytogether permit a major reductionin the frequency of errors.But both those errorsthat subsistand the new sorts of errorproper to such systemshave a greateraverage unit cost, due to theirinstantaneous propagation, their inherent complexity, their unpredictableform, and theircorrespondingly increased discovery and rectificationcosts. The total cost of errorswill normallybe reduced, but the nature of the quality control problem is transformed.9

4. Skill Impact of Banking Automation

How can we characterizethe correspondingshift in skill requirements?At thispreliminary stage, it is usefulto cast our net as wide as possible, and thereforeto begin with as pragmatican approachas possible.

I have adopted a frameworkderived from job evaluation practices, a relatively standard twelve-categorybreakdown of distinct"worker contributions," that was used in a land-markstudy of automationby JamesBright (1958a). This sectionanalyzes each of these contributionsas they have evolved in the course of the computerizationof low-level clerical operationsin banking,and comparesthe latestphase of bankingcomputerization with Bright's generalizationsconcerning the higher levels of automation.The interestof this comparisonis enhancedby the factthat Braverman (in Ch. 9) relies on Bright's analysis as one of the major exhibits forhis deskillingcase.

The use of banking as an example of a highly-automated process is perhaps surprising,but entirely appropriate. The distinguishingfeatures of the highest levels of automation as definedby Bright-and the definitionhas dated surprisinglylittle- are these: thatlong sequences of operationsare conductedwithout any human intervention(in banking: automatic updating of multiple interrelatedaccounts), except for the input of raw materials(data) and a constantsurveillance; the systemcontrols for the qualityof inputs(by computerchecks for data plausibility

8. In reality,some accountsare changedso oftenthat continuous updatingof the permanentfile is too expensive.These data are fed automaticallyto locally processed,provisional files, which are then automaticallyfed into central permanent files at theend of theday. 9. For an identicalgeneralization derived in an altogetherdifferent context,see Weiner,1985. 10 BERKELEY JOURNAL OF SOCIOLOGY

and consistency) and for the compatibilityof its own sub- operations (by account balancing tests); the system anticipates action requiredand adjusts to provideit (with advanced computer systems equipped with dynamic processing,multiprogramming, and virtualmemory.)

Table 1: JamesBlight's assessmentof the skill requirementsof the highestautomation level. WorkerContribution or SacrificeTraditionally Requirementat Highest Receiving Compensation Level of Automation

Physical Effort Nil

Mental Effort Decreasing-Nil

ManipulativeSkill Nil (dexterity)

General Skill Decreasing-Nil

Education Increasingor Decreasing

Experience Decreasing-Nil

Exposure to Hazards Nil

Acceptance of Undesirable Decreasing-Nil Job Conditions

Responsibility" Increasing,Decreasing or Nil

Decision Making Decreasing-Nil

Influenceon Productivity*** Nil

Seniority Not Affected

"Refersto operatorsand not to setup or maintenanceworkers, engineers, or supervisors. **Safetyof equipment,of the product,of otherpeople. ""Refersto opportunityfor the workerto increaseoutput through extra effort,skill, or judgment.

Source: Bright,1958b.

The conclusion Bright drew concerning the evolution of skill requirementsin automated contexts is summarized in Table 1.

Bright's is a largely pessimistic diagnosis from the workers' point of view.10 Only "Education" and "Responsibility"

10. The one positiveelement in Bright's analysis fromthe workers' point of view is that he saw little probabilityof large numbers of workersfinding themselves so undertrainedas to be unemployable.For ADLER: AUTOMATION AND CAPITALISM 11 requirementsare not forecast in regression-and then only provisionally,since Bright (1958b) writes: "When a patternof predeterminedactions can be mechanicallyachieved, there is no particular need for the understanding,the training, and the educationon the part of the operator"and "the ultimateeffect of the higher levels of automationis to remove responsibilityfor performancefrom the hands of the worker."

Let us examine the banking case factor by factor and compareit withBright' s assessment.

(1) Physical effort

Bright's assessmentof decliningphysical effortrequirements seems difficult to dispute. He underestimatesthe various ergonomieproblems-which, it is true,have only recentlycome into prominence.It is unclear under which rubricthese problems shouldbe classed; I shall leave themto the "Hazards" section.

(2) Mental effort

Bright's diagnosis of reduced mental effort reveals an importantshortcoming in the application of his framework.In relationto the highestlevels of automation,he writes (1958b): "By definition,the more automatic machines employ control devices thatregulate their performance to achieve the desired end withouthuman attention.Therefore, mental strainas a result of mentaleffort is ultimatelyreduced."

The automatic control capacity of the advanced machine systemmay well reduce quantitativelythe human's role; such is the basis for Bright's argument.But Bright seems to be so impressedwith automation'spotential for self-regulationthat he ignores the original question: given that employmentlevels will adjust to reduce any redundanciesin the work force,what is the qualitativenature of the remainingjobs?11

In banking,the increasingdegree of automationmeans the replacement of a sequence of specialized departmentseach

managementon the otherhand, Blight's prognosiswas overwhelmingly optimistic,since the predominantfear at the time of his research - the 1950s - was that automationwould require massive and expensive retrainingprograms.

11. In fairness to Bright, it should be recalled that he was interveningin a context dominatedby the fear of massive short-term increases in skilled-laborrequirements. He was overstatinghis case on this subject when he venturedinto the broader,longer-run issues we are exploring. 12 BERKELEY JOURNAL OF SOCIOLOGY performinga small numberof elementarydata-entry tasks by a systemin which a single operatorenters into the systemall the variables needed for each transaction.A simple operation like cashing a check involves 19 distincttasks of several different kinds (data-entry,data verification,terminal procedural commands, physical manipulation)that the teller, workingwith an on-line terminal,must execute in precise order,respecting the software's sequence of decision branches.Each separatetask is simple, but theircorrect combination and ordering,especially while scrutinized by a line of impatientcustomers, demands the internalizationof a new sortof intellectualdiscipline: the abilityto masterprocedural, algorithmicmodes of thought(Shiel, 1981). We shall returnto the impact of this aspect of automationon other factors,but we should note here that the mental effortinvolved in constantly maintaininga "clear head" is not to be underestimated.It is a new formof effortfor many workers.

Another feature of particularimportance is the fact the interaction with the automatic system (as with advanced manufacturingequipment) is mediatedby a multiplicityof codes. Even when softwarebecomes more the user-friendly, " operator- machine interfaceis not one characterizedby truly natural" language. A common illusion here is thathabit can renderbanal the new code-mediatedtasks. Detailed research has shown that even after a long ,coded data-entryinvolves perceptualprocesses thatare economically more demandingthan those involved in the entryof regular text; whence a second sourceof mentaleffort Bright ignores (Durraffourg et al., 1979).

To step outside the sphereof low-level bank operatorsfor a moment,we should note in passing that there is another,less obvious, form of mental effortthat Bright underestimates.In discussing the highest automationlevels, Bright writes: "The progressiveeffect of automationis firstto relieve the operatorof manual effortand then to relieve him of the need to apply continuousmental effort" (1958a, p. 188). He describesthe role of the operator of more advanced machineryas "patrolling"or "machine-tending,"evoking the image of the night watchman. Studies of refineryworkers indicate, however, that the mental effortof surveillanceimposes a considerableburden of a new kind, when continuousconcentration of the kind associated with manual fabricationactivities is replaced by the strainof continual watchfulnessand readiness (see Galle and Vatin, 1980 and Edwards and Lees, 1974).

Highly automated systems generate a new and complex operatingmode which demandsof workersthat they be ready to react rapidly even afterhours of apparentinactivity. Associated with this new operatingmode is a new sort of straindue to the conjunctionof (a) high levels of responsibility(to which subject ADLER: AUTOMATION AND CAPITALISM 13 we shall return),and (b) modes of system failure increasingly difficultto predict(on thislast factor,see Hirschhorn,1981).

Brightignores such problemsin his faiththat machines will eventuallycontrol themselves.But as control functionsbecome more sophisticated,the problemis only removedto higherlevels of responsibility.

(3) Manipulative skill (dexterity)

Here, as withphysical effort, it is clearlydifficult to dispute Bright's claim that dexterity requirementswill decline as automationlevels rise. Typingskills have become a more common requirement,but in theirspecifically manual componentthese are of a lower level of complexitythan many of the manual skills automationdisplaces. (4) General Skill

It is of particularinterest to findthat the case of bankingruns counterto Bright's diagnosis of fallinggeneral skill requirements as measuredby trainingtime.

Training programs are expensive, and can be cited as justificationfor higherpay scales. But the managersI interviewed expressed their concern that underestimatingthe new training needs would prove to be myopic,by limitingthe capabilityof operatorsnot only to fullymaster their current tasks, but also, and perhapsmore importantly,by hamperingtheir ability to adapt to constantlyevolving computer procedures.

Bright (1958a) argues: "In many instances, the need for educationand understandingof principlesmay continuewell into the higher levels [of automation].However, these eventually become unnecessarycontributions as reliabilityincreases." So imagine the bank managers I interviewed-before the computerizationprocess moved into the highest levels of automation.

The managersresponsible for computerizationwere haunted by the followingscenario: the client comes in, angry,wanting to know why the transfers/he effecteda monthago has not been creditedto his/heraccount. The tellerhas no one to turnto. It was indeed the tellerwho had effectedthe operationon the terminalat the frontdesk. There is no longer a chain of processingoffices handlingthese operations.By the same token,there is no longera series of office chiefs to whom the teller can turn in order to determinewhere along the processing chain the operation has brokendown. 14 BERKELEY JOURNAL OF SOCIOLOGY

Hence the "bankers' paradox": they had indeed hoped to simplifyoperations and thereforereduce trainingrequirements when they computerized.Despite the public presentationsabout enrichingteller jobs, such was in fact theirmain concern. And now, committedto a vast programof branchcomputerization, they discovered-with not littleanxiety-the magnitude of the training task beforethem. If a modicumof what they have come to call "local mastery"were not sustainedat the employeelevel, the bank was liable to finditself swamped by angrycustomers. The lesson of the banking case is straightforward:the need for "local mastery"grows as the level of automationincreases.

The only possible responseto this pressingneed is training. Three areas of trainingdevelopment have emergedas particularly important: (a) basic computerliteracy; (b) informationregarding the structureof the bank's processingsystem; (c) and, for(b) to have any meaningfor the employees,a muchbroader understanding of the logic of bank accounting.

The bank I studiedmost closely, like othersin France, was thereforebusy developingeducational courses, software, and video cassettes.It was still unclearhow many hours of trainingwould be necessary.But banks, because of theirline of business,were not, unlike many otherindustries, able to burytheir heads in the sand and ignorethe trainingimperative associated with all but the technicallymost elementaryjobs: traininghas had to go beyond the immediatelyfunctional if it is to be reallyoperational. (5) General education

Clearly, sooner or later some of the basic computerliteracy describedin the previousparagraph will come to be accomplished in the frameworkof generaleducation. This shiftand upgradingin skill requirementsexplains the shiftin recruitmentcriteria for entry-levelpositions to a minimumof a high school diploma and a preferencefor two years of college (Cossalter and Hezard, 1983).

It is, furthermore,not just the quantityof educationdemanded thatis changing.It is also its contentand formthat evolve. The "clear thinking" requirementsof the associated procedural operatingmodes seem to pose a challenge to the educational system.

12. We are witnessinga culturalmutation of considerable proportions,perhaps to be comparedto the "Americanization"process livedby thewave of immigrantsthat populated the Ford assembly line at ADLER: AUTOMATION AND CAPITALISM 15

(6) Experience

Bright's diagnosis-declining experience requirements-is here also dictatedby his assumptionthat the machine will eventually be totallyreliable in doing what the workeronce did. Hopefully my pointon thisscore has alreadybeen made. But we still need to characterizethe absolute level of experiencerequired as well as the changingrelative weights of education,training and experiencerequirements as the level of automationrises.

To take the relativequestion first: to the extentthat , numeracyand technologicalculture requirements rise, one could hypothesizethat experience-based "tricks of the trade"will play a correspondinglyreduced relativerole. The bankingcase supports this hypothesis.I noted above the need for basic accounting trainingfor low-level bank personnel.Pushed as they are to the peripheryof the accountfabrication process, their mastery of their own tasks demands some understandingnot just of the computer systemin its technicaldimension, but also of the accountinglogic which governs the inter-relationof the various types of bank accounts. In the previous stages of bank technique, no such breadthof trainingwas required.All the functionaltraining for low-level positions was acquired on-the-job. This no longer suffices.

The problem is posed even more acutely for positions requiring greater technical expertise and/or responsibility. Previously,it was by the successive apprenticeshipof several departmentsthat upwardly-mobile personnel learned their banking. Now, experiencecannot suffice,as critical operatingprocedures have been internalizedin the computersystem.

In both instances,for operating personnel and forhigher-level personnel,education and trainingrequirements are increased,and on-the-job experience requirements become relatively less important.

On the otherhand, computersystems, despite being situated the beginningof the century.Whereas Ford's problemwas developing body disciplinesand life-stylesthat would permitimmigrants from largelyrural backgroundsto operateassembly-line facilities, today's problemis expressedin a New YorkTimes article this way: "Educators are startingto ask whetherthe way Americansthink, the way theygo aboutdefining and solvingproblems, will be altered,for better or for worse,by increasingexperience with the systematicand quantitative thinkingthat goes intoprogramming a computer" (Fiske, 1982). 16 BERKELEY JOURNAL OF SOCIOLOGY at veryhigh levels of automation,are not entirelystandardized- far fromit. If small-scaleusers usuallybuy packaged software,larger users are constantlyadapting their applications. The resultis that many operatorshave to learn quite idiosyncraticcompany-specific systemsin constantevolution.

The overall effectis thatthe absolute level of the experience requirementoften increases with automation,even though its relativeimportance decreases.

(7) Exposure to hazards

It seems reasonable to hypothesize that with the "peripherization"of workersvis-a-vis direct product manipulation, workplacehazards are probablyas a generalrule on the decline. It seems reasonable, too, to note that the harsh conditions of yesteryearare easily forgotten,magnifying the relativelyminor inconveniencesof today. But thereare some ergonomieproblems associated with currentcomputerized equipment, problems that cannot easily be relegated to the status of "undesirablejob conditions."

Currentcomputer display systems can aggravate eye- and back-strain.The problem is partlyin the equipmentdesign and installation.Too often no provision is made for adjustmentof lighting, height, tilt, physical configurationof screen and keyboard,etc., violatingelementary industrial engineering norms. For anotherpart, the problemresides in the currenttechnology of cathode-ray tubes: limited definition,high scintillation,poor contrastcontrol, etc.13

Clearly, improvementsin this area are needed. Unrectified, such strain can lead to serious injury. Equally clearly, such problems are not inherentin the automation level; they are surmountable,if only underthe impetusof organizedintervention. Some Frenchbanks are alreadyusing liquid crystaldisplays which dramaticallyreduce most of the visual problems. The adoption was motivatedby management'sdesire to still union and public criticism.

The key ergonomieproblems, however, seem to be due not to the level of technology,but to the organizationof work. High- concentrationtasks with highly constrained operating modes, exemplified in such jobs as data-entryclerks, should, for ergonomiereasons, not be maintainedfor more than a couple of hoursat a time (cf. Grandjeanand Vigliani, 1980). Such has been

13. See the various publicationsof the National Instituteof OccupationalSafety and Health. ADLER: AUTOMATION AND CAPITALISM 17 the recommendationof various European bodies, and more and more employersare respondingto the challenge of this normby de-specialization,incorporating data-entry tasks into other jobs, like the bank teller's.

Thus, while not minimizingequipment-related problems, I would hypothesizethat not a few of the ergonomie problems ascribed to new technologies reflect a displacement onto technologyof frustrationwith other factors, like workorganization or job content.Indeed, the next paragraphshows that the very natureof automatedwork may be aggravatingthese other factors.

(8) Undesirable job conditions

To observethe bankingindustry is to witnessa strangescene. Effortrequirements may be on the decline; skill requirements seem to be on the increase; yet jobs are not experiencedas any more satisfying.On the contrary,the lack of interest that characterizesmany banking jobs is eventuallyclassified as an "undesirablejob condition"in itself. An example: French banks have been forcedto give a high priorityto automatingsecurities departmentprocessing jobs for the simple reason that these jobs were experiencedas so boring that stable staffingcould not be assured.

For the mass of workerswho will assure the interfaceand surveillance functions,like the bank employees in their vast majority,the scenarioseems to be this: as the fabricationprocess is automated,as the workeris renderedincreasingly peripheral to it, the operator'swork is reduced fromits concrete,tangible form to a more abstract activity-interface and surveillance. In this transformation,the concretenessof the task's goal is diminished. There is clearly a segment of the work force for whom the "concreteobject" is the systemitself: programmers, managers, etc. Their case is more complex. But most employees find that in place of the easily apprehendedgoal of correctlyeffecting a distinctoperation, the new situation substitutesa much more amorphousobjective-that of ensuringthe smooth functioningof an almosttotally automated set of interrelatedoperations. With the peripherizationof the worker,it is not only the meúW5-computer- mediatedtransactions via coded commands-butalso the ends of workthat appear as less concreteand more abstract.

If the abstractionof means increases trainingrequirements and mentaleffort, the abstractionof ends leads to a very different result-workis oftenexperienced as intenselyboring.

(9) Responsibility

Despite Bright's agnosticismand (ultimately)pessimism on 18 BERKELEY JOURNAL OF SOCIOLOGY this factor's evolution,the evidence frombanking, as well as a closer readingof Bright's own refinerycase, would suggestfairly unambiguously that responsibility generally increases with automation.

The key considerationin the bankingcase is the multiplicity of operationsthat the computerconducts on the basis of a single data-entry:the teller's operation,sometimes augmented by thatof the remainingprocessing units, feeds a multiplicityof accounts at all levels of the bank's records.The banks have thereforeinvested considerable effort in developing data consistency/plausibility checks. These, however,are neverperfect. As infrequentas errors may thus become, the complexityand thereforethe unit cost of the remainingerrors are correspondinglyincreased.

The concernexpressed by bank managersfor the maintenance of "local mastery"is rooted in such problems. Similarly,the reorganizationof the remainingprocessing departments into zones of geographicresponsibility14 tears witnessto the operational,and not simplyhumanistic ideological, motivation for this concern with responsibility. The same motivation encourages banks to reorganizeinto more autonomouswork teams to allow employees to assume thisbroader responsibility.

The importof such an increase in responsibilityfor even banal bankingjobs is, I believe, considerable.The centralissue is thatthe responsibilitysolicited from workers is no longerconfined to the object of maintainingwork-effort standards. The workers are asked to play a differentrole: theyare asked to be responsible for the integrityof the process and for its results. Of course, managershave always wantedas much. What has changed is that now the technical characteristicsof production make such responsibilityan operationalimperative.

(10) Decision making

This factor is closely linked to the preceding one. The difficultyof any diagnosis here is that this factor conflates changes in the scope of decisions with changes in the degree of authorityin decisionsof a given scope.

The banking case lends some support to the thesis that authoritycan be centralizedin a set of operatingprocedures writteninto the software,thereby reducing the autonomy of decision of the local personnel. Such centralization,however, becomes costly-for example, in poor quality results-when it

14. Citibankhas undertakena similarreorganization: see Matteis, 1979. ADLER: AUTOMATION AND CAPITALISM 19 undercutsthe real masteryof local operations.Centralization of key (strategic) decisions is thus commonly combined with a change in the scope of decisions, to encourage a concomitant decentralizationof minor(operating) decisions.

As othershave also pointedout, a significantfeature of the new configurationis that organizationalcoordination structures themselves become more "abstract" insofar as they are increasinglyembodied in softwareroutines rather than in the personal authorityof the local management(Zuboff, 1982). This automationof coordinationis not withoutdramatic repercussions on the supervisor'srole. Technologicaltrends seem to encourage the shiftin the sources of supervisors'legitimacy from technical expertiseto team facilitator.15

To conclude, however,as Brightdoes that decision-making contributionsof operatives are decreasing or nil, is manifestly inadequate.

(11) Influenceon productivity

I shall not restate my case for believing that worker performancein automatedenvironments is decisive foroperational efficiency.Blight's positiondoes, however,reflect more profound featuresof the situationthat indeed meritattention.

Under automatedconditions, major productivityincreases can no longercome fromrationalizing the worker'sgestures, as in the Taylorist mode. Freeing the production process from its dependence on the craftsman's dexterityand eliminating the technical inadequacies that created numerous "pores" in the workingday have bothfacilitated intensification of workeffort.

Meanwhile, behind the scenes as it were, the progressive mechanizationand automationof production,and the economic rationalityand operationsrationale that have developed with it, have been workingto attributeto human interventiona different type of efficacy.Science and technologyprogressively displace both dexterityand intensityof effortas the key to productivity increase.This tendencyreinforces the significanceof the worker's role in guaranteeingthe smooth functioningof the highly automatedsystem.

15. This shiftgenerates new legitimacyproblems. The criteriafor promotionto supervisorlevel give less weightto experienceand technicalexpertise, and more weightto personalityqualities. More subjectivecriteria mean more fragile authority. 20 BERKELEY JOURNAL OF SOCIOLOGY

Table 2: Comparingthe BankingCase and Blight's Assessmentat the HighestAutomation Level

BankingCases Worker Blight's Quantitative Qualitative Contribution Assessment Changes Changes 1) Physicaleffort Nil Decreasing But see Hazardson strain

2) Mentaleffort Decreasing Increasing Algorithmic -NU operatingmodes desiplacemanual gesture 3) Manipulative Nil Decreasing But: see Experience re: "tricksof the trade"

4) Generalskill Decreasing Increasing Technological -NU literacy Local Mastery CognitiveLearning 5) Generaleducation Increasingor Increasing Technologicalculture Decreasing or Decreasing Intellectual discipline("clear head") 6) Experience Decreasing Decreasing Decreasingrelative to -NU and/or education,but Increasing increasingabsolute leveldue to dynamic large-system idiosyncracies 7) Exposureto Nu Decreasing But:pervasive strain hazards withhighly constrainedoperating modes

8) Acceptanceof Decreasing Decreasing Intrinsicboredom undesirablejob -NU objectively,but displacesphysical conditions Increasingsubjectivly conditions as key issuesubjectively 9) Responsibility Increasing, Increasing Responsibilityfor Decreasingor NU production,not just effort Teameffort, not just individualeffort

10) Decisionmaking Decreasingor NU Decreasingand/or Centralizationof Increasing parameters Decentralizationof operatingvariables Note:more abstract decisionprocesses 11) Influenceon Nil Increasing Costof errorsand productivity downtime(work efficacy,not intensity) Role interdépendance 12) Seniority Notaffected Decreasing See: Experience and/or Increasing ADLER: AUTOMATION AND CAPITALISM 21

One source of Bright's confusionmay lie in the factthat as a generalrule the worker'scontribution to productivityis much less individual in more automated settings. The effectivenessof automatedsystems is governedby the quality of the interaction between members of the work team, between technicians and workers,between Engineeringand Operations,between all these and the Training department,and between manager and employees.

(12) Seniority

This factoris often a surrogatefor experience.As such, I have suggested that its importance is subject to conflicting pressures,and may, as a result,be decliningrelative to education and trainingbut increasingin absoluteterms. A furtherdownward on this factor's importance,perhaps a transitoryone, but one currentlycausing significanthardship, is the perceptionof greater flexibilityof youngerworkers compared to older workers.Older workersare less likelyto have the requisitenew skills, and older workersare oftenpresumed-often incorrectly- to be less able to adapt. 5. From "Worker Contributions"to Dimensions of Skill

In summary,we can compare Bright's evaluation of the impact of automationon workercontributions with the preceding analysis of banking.Table 2 recapitulatesthe salient featuresof both evaluations,comparing Bright's prognosisof the skill shifts associated withadvanced automationand my analysisof advanced computerizationin banking. Examination of this table leads me to two principle conclusions.First, in contrastwith Bright, the bankingcase shows at least as many quantitativeincreases as decreases. Notice, furthermore,how on some of the contributionswhich have assumed particularimportance in my analysis, Bright is either undecided and tending to pessimism (general education, responsibility)or is in starkdisagreement with the analysis of the banking case (mental effort, general skill, influence on productivity).

More importantperhaps is that my rejection of Bright's pessimism is reinforcedby considerationof the changes in the relative weightsof the differentworker contributions. From this point of view, most of the decreases we observe can only be judged welcome: the decrease in major hazards,the substitutionof mental for physical effort,of general skill and education for manipulativeskill, experience, and seniority.

The second principleconclusion I draw fromTable 2 is even more criticalfor the argumentof this paper. Reviewing the last 22 BERKELEY JOURNAL OF SOCIOLOGY column of Table 2, we can identifythe symptomsof a misfit between Bright's twelve-dimensionalanalysis and the observed evolutionof work. This misfitleads me to suggest that beneath Bright's structurethere is a deeper, four-dimensionalstructure at work: (1) degree of substantivecomplexity (low to high); (2) type of responsibility(responsibility for effort versus responsibility for results); (3) the type of expertise(manual versus cognitive),and (4) type of interdependence(individual or sequential versus collective). The three qualitativedimensions are summarizedin Table 3.

Table 3: Threeunderlying qualitative dimensions of work QualitativeChanges Dimensions - Local mastery Social responsibilityfor process - Responsibilityfor integrityand results production versus - Influenceon productivity - Costof errors Private purelyinstrumental attitude to work,responsible primarilyfor work effort -Algorithmicoperating modes Abstract abstract,mental, machine- -Cognitivelearning mediatedtasks with -Decision-makingprocesses cognitvelearning but with -Newergonomie problems versus low inherentinterest -Computerliteracy -Importanceof generaltraining -Intrinsicboredom Concrete tangibleimmediacy of -Intellectualdiscipline tasksand goals,with manualor rotelearning -Teamresponsibility Collective systematicinterdependence -Roleinterdependence criticalrole of coordination betweentasks and between versus departments;job rotation andteamwork encouraged Individual stand-aloneor purely mechanical,sequential linkageof individualized jobs

Two claims can be made for this four-dimensional framework.

First, a strongclaim: the banking case suggests that these four dimensionsare ones that managementhas to address in the medium-longrun in designingjobs. Any theoreticalanalysis of skills should thereforeencompass them. ADLER: AUTOMATION AND CAPITALISM 23

Evidence fromother industries supports the bankingcase in this.Job evaluationsystems, for example,obviously deal with the first,quantitative dimension, but have also been shown to be sensitiveto the otherthree, more qualitative factors:

-If the benchmarkjobs are characterizedby responsibility-for- effort,responsibility-for-results will be undervalued. The steel industryhad to confrontthis issue when it became obvious that production operators' responsibility for assuring operations continuity was a greater productive contributionthan the craftmen'straditional responsibility for careful parts machining (Steiber,1959). -If the benchmarkjobs are all "concrete,"the system will undervaluemore abstractforms of expertise,particularly in their effortand general skill contributions.Women's clerical jobs have oftenbeen undervaluedin this way, the menial physical exertion of many "male" jobs oftenbeing privilegedover routinemental effort(Treiman and Hartmann,1981).

-And if the systeminterdependence is such that a sustained integrationof learning and doing seems necessary, the job evaluationapproach itself niäy have to be jettisonedor profoundly modified in favor of a system in which wages reflect skills acquired rather than the requirementsof the particular job currentlyoccupied (Lawler, 1981).

Cain and Treiman's (1981) statisticalanalysis can also be marshalled to provide surprisinglystrong support for this framework.16They performeda factoranalysis on the 44 variables used by the Dictionary of Occupational Titles to characterize occupations.Factor analyzinga 10% sample of the DOT 's 12,099 occupations,they found factorswhich correspondclosely to the dimensionsI have identified:

(a) a "substantivecomplexity" factor captures the general trainingrequirements of the firstquantitative dimension;

(b) "motor skills" and "physical demands" factors are summarizedin my manualversus cognitive expertise;

(c) a "management"factor captures some of the elementsof the responsibilitydimension;

(d) an "interpersonalskills" factorcaptures-as best the DOT

16. Surprising,that is, to the author.The analysisof Cain and Treimanonly came to myattention some time after this framework had beenarticulated. 24 BERKELEY JOURNAL OF SOCIOLOGY characteristicsallow- the interdependencedimension;

(Cain and Treiman also identified an undesirable working conditionsfactor that is irrelevantfor our purposes.) Evidence of the broadly positive natureof this shiftin the banking case is indirectbut eloquent:between 1969 and 1977, the period of most intensivecomputerization, the proportionof lower-tieremployees in Frenchbanks fell from58% to 45%, withcomplementary gains in the category representing more-skilled employees and supervisors (32% to 42%), and to a lesser extent in the professionaland managementcategory (9% to 13%).17

In one of the largest banks (approximately 40,000 employees),the evolutionof the personnelstructure demonstrates this upgradingover an even longer period (Delegation Nationale CFDT, 1985). Their detailed statistics make it possible to distinguishwithin the middle-tierthose withoutany supervisory responsibility(at least as of 1970, when the middle-tierwas redefinedto includea firstlevel withoutsupervisory responsibility. Such a promotion opportunitybegan to appear necessary to accomodate the upgradingof the employee skill base). Table 4 shows a dramaticupgrading trend. Table 4: Structureof personnelin one majorFrench bank. Year Lower-tierMiddle-tier Employees ManagementTotal Employees and Without With Professionals Supervisory Supervisory ResponsibilitiesResponsibilities 1950 74.4% 0 17.9% 7.7% 100% 1960 68.2% 0 22.1% 9.7% 100% 1970 59.8% 11.6% 19.2% 9.4% 100% 1980 35.2% 25.9% 26.4% 12.5% 100% 1983 31.0% 28.5% 27.6% 12.9% 100%

Source:Delegation Nationale CFDT, 1985.

Follow-up interviewswith personnel directorsin this bank (conducted in June 1985) revealed a broad consensus that such shiftsdid indeed representskill upgrading,and thatthis upgrading trend has continued Task complexity has increased, if only modestly, and there has been signigicant horizontal job "enlargement"and vertical job "enrichment":the level of

17. Data suppliedby theAssociation Française des Banques. ADLER: AUTOMATION AND CAPITALISM 25 responsibilityrequired of operatingpersonnel, the breadthof these responsiblities,and the need to understandbroader segmentsof bank operationshave been major preoccupationsof trainingand personnelmanagers.

The generalizabilityof these results is naturallysubject to debate. I have elsewhere(Adler, 1987) revieweda series of recent case studies which suggestthat upgradingmay be more frequent than the researchconducted in the wake of Bravermanmay have led us to believe. More importantly,Spenner's review (1983) of the major statisticalstudies conductedto date shows that of the seven studiesof job content,five showed upgrading,two showed no effect,and none showed deskilling,and of the six studies of labor force composition,five showed upgrading,one showed no net effect, and none showed deskilling. The case for net upgrading-albeitmodest- for both occupationsand the labor force as a whole is much strongerthan many have suggested.

The bankingcase cannot by itselfprove, but it can certainly suggestthat, as a generalrule, and despitea greatdeal of diversity in the situations of firms and occupations, automation will encourage shifts in a distinct direction along each of these dimensions,towards:

- more substantivecomplexity, - moreresponsibility for results, - more abstractreasoning, and - morecomplex interdependence.

Certainly,many other factorsapart fromtechnology bear on skill levels and job content;and indeed many otherfactors apart fromoperations efficiency bear on the choice of technology.But the banking example, like many others read through these theoreticallenses and like the admittedlyweak statisticaldata available, suggests that deskilling and degradation of work- interpretedas Braverman's "secular trend toward the incessant loweringof the workingclass as a whole"-may be a myth.

6. The Dimensions of Skill and the CommodityForm

The key insightderived fromthe banking example is in the idea that any analysis of skill and automation will have to confrontboth the quantitativedimension of substantivecomplexity as well as the more qualitative dimensions of responsibility, abstractnessand interdependence. These termsreflect the intersectionof the inductivemoment of my researchand a deductiveexperiment conducted in parallel. I should thereforeexplain the natureof this deductiveexperiment and readerswill be able to inferfor themselves how successfulthe 26 BERKELEY JOURNAL OF SOCIOLOGY overallexperiment has been.

The experimentconsists in taking Marx's conceptionof the commodityform as a guide to the analysis of the dimensionsof skill. Brieflyput, the substantivecomplexity dimension represents the value of relativeskill as measuredby trainingtime. The three qualitativedimensions represent the labor market'sthree principal conditionsof existenceas a means of governingthe production and allocationof the commoditylabor-power. These conditionsof existencecan be characterizedby threepolarities:

- Social/private: The needs of societyare satisfiedby the sale of productsproduced in independent,private enterprises. There is a basic split between the social natureof those needs and the privateownership of the firmsthat are the means of satisfying these needs. This split is the most fundamentalcondition of any market'sexistence, including the labor market:supply and demand meet on the marketplace, not in some preestablishedeconomic plan.

- Abstract!concrete: The exchanges that take place between buyers and sellers are exchanges that are normed by the equivalence of the quantityof abstractlabor embodied in the commoditiesbeing boughtand sold. Abstractlabor is labor made commensurableby a process whereby we abstract from its particularconcrete qualities.

- Collective/individual: Even thoughthe laborer,the bearer of labor-power,is a "free individual" and the exchange of labor- power forwages is an individualizedtransaction, the labor process is basically a collectiveactivity, drawing many individual workers intoan enterpriseunder common direction.

The deductiveexperiment consists in takingthese theoretical and macro-socialconditions of existenceof the labor marketand effectinga double shiftby findingfor them corollariesthat are both concrete (as opposed to theoretical)and centered in the "micro"reality of the labor process (as opposed to referringonly to macro-societalphenomena):

- social/private: we can interpretthe shiftin the nature of responsibilityrequirements as expressinga "socialization"of the worker's role, since with automationthe workeris increasingly asked to contributenot only his/hereffort but also his/hersense of responsibilityfor the integrityof the process and forthe qualityof the product;the workerno longerappears in the labor process as a mere homo economicus,but instead as a being with social as well as economicdimensions.

- abstract/concrete: we can relate the theoreticalcategories ADLER: AUTOMATION AND CAPITALISM 27

pertainingto the commensurabilityconditions of existence of the market to the abstract or concrete nature of tasks, to the correspondingcognitive or manual-rotenature of expertise,and to the theoreticalor experientiallearning process.

- collective/individual:this polaritycan be applied directlyto the natureand degreeof interdependencebetween workers. The idea of such a theoretical-to-concreteshift comes from Marx.18What I have done in identifyingmy three qualitative dimensionsof work is to pursuethis operationof concretionfrom the more macro level, at which Marx situateshis commentson socialization and abstraction,into the more micro level of the labor process itself. Let us thereforereview the four dimensionsof skill in this light.In each case, the interpretationof the concept in a concrete and micro form uncovers a basic contradictionbetween the productivity- and technologically-driven requirements of contemporaryindustry and traditional personnel managment practices which reflectthe commoditystatus of labor. In each dimension,we shall be able to highlightautomation-induced trends in work which both underminethe commodityform of labor and suggest the emergence of new, post-capitalist forms of organization.

18. Marx (1973, pp.101-104). Marx argued that the reason his concept of abstractlabor is pertinentis thatlabor had in realitybecome abstract.By this he refersto the breakdownof secrets and to the mobilityof workersbetween differentlines of work - which are the concreteconditions making possible the comparisonof labor times. Prior to such labor mobility,real labor times were a closely guarded secret of the , and abstractlabor was thereforea purelytheoretical construct. Similarly, Marx argues that the social character of production is evidencednot only in the somewhattheoretical fact of privateproduction for social needs but also in somethingmuch more tangible,namely, the interdependenceof activitiesin differentproducing units. Finally,Marx's conceptof the collectiveworker is not only an abstractconcept designed to reflectthe fact thatmost productiontakes place in plantsthat employ more than one person, but is also a concept that permits concrete analysis of specific structuresof cooperationthat characterizedifferent industrialforms. In this way, Marx establishesthe foundationsfor a set of what he calls "concreteuniversais." This somewhatpeculiar concept emergesin Marx's discussionof abstractlabor in the firstGerman edition of Capital, where he makes the apparentlypreposterous claim that the real world operates as if, alongside all of the animals in the animal kingdom,there existed as well an animal called "Animal" (Marx, 1867, para. 65). In otherwords, alongside the various formsof concretelabor thereexisted somethingcalled "abstractlabor" thatwas just as real, just as tangible,and just as importantin its effects. 28 BERKELEY JOURNAL OF SOCIOLOGY

In this way, we returnto Braverman'schallenge of linking labor process analysis to broader societal themes. But in my analysis, we will find a progressive strengtheningof worker capabilities,not a processof degradation.

(1) Complexity

That automationtends, on average and over large aggregates, to generateincreases in the "substantivecomplexity" of work is well enoughaccepted (Spenner,1983). But whethersuch increases underminein any directway the commodityform of labor is less obvious: of the fourdimensions I have identified,this one, being quantitative,is the one most directly linked to labor as a commodity.My principalargument will be that changes in the otherdimensions of skill underminethe salience of this dimension: as work becomes more responsible,abstract and interdependent, either the trainingtime required to engender these qualities becomes impossible to measure, or that measure becomes ineffectivein assuringwage equity.

But a secondaryargument is also in order: as the level of complexityrises, we should expect greater firm-specificityof skills. This correlationis not verytight, since high-levelpersonnel in some specialties are notoriouslyfoot-loose. But as a general rule, and one subjectto empiricaltesting, we could expect thatthe greaterthe complexityof the individual's knowledge-base,the more numerousthe potentiallyvaluable tacitlinkages between that base and the contextof its deployment.

If the firm-specificityof skill increases, the normal labor marketprocesses fail, and non-marketallocation processes internal to the firmemerge (Doeringer and Piore, 1971). Many studentsof so-called "internal labor markets" have presented the basic rationalefor such institutionsin the somewhatconspiratorial terms of "divide and conquer" (Gordon, Edwards, and Reich, 1981). Othersnote these arrangements'economic efficiency under a set of assumptionssomewhat more plausible thanthose of the traditional neoclassicaltheory of the firm(Yellen, 1984).

But a verydifferent reading of the issue is possible,focused on the fundamentalhistorical significance of the shiftfrom a market to a non-marketmode of labor allocationfor a growingproportion of the labor force.

Edwards (1979) presentsa fascinatingsketch of the evolution of labor control systems from simple (supervisory)control to technical(technology-mediated) control to bureaucraticcontrol. He analyzes this sequence as a series of independentchanges brought about by class conflict.If the sequence has any historical"sense" for him it is that it has been a process which "turnedconflict ADLER: AUTOMATION AND CAPITALISM 29 withinthe firmdecisively" in favorof capitalists(p.200). But the same evidence can be read~and I believe, read more plausibly-as a sequence expressingreal socialization,as capitalistscede their authorityin turn to norms of technical efficiencyand then to negotiatedpolitical/ethical norms of fairness,rights, and duties.

The pervasivenessof such a shiftis indicatedby the frequency with which even non-unionized"progressive" firms impose on themselvesstrict ethical norms and administrativeprocedures in orderto maintaintheir workers' confidence.

The developmentof internallabor nonmarketsthus tends to underminethe viabilityof the commodityform of labor at the same time as it tends to establisha new labor allocationprocess based on negotiatedauthority.

(2) Responsibility

As automationlevels rise, I have suggestedthat responsibility- for-efforttends to become responsibility-for-results,and the "managerial"content of work thus increases. The workeris no longermerely the one-dimensionalhomo economicusof the wage- effortbargain, but is called upon to take an active role in assuring the success of a collaborativeventure. This representswhat can be construedas a socialization of the worker's role in the labor process-adding social depth to the one-dimensionalityof the wage-laboreras homo economicus.

Responsibilityappears to be movingin this more "socialized" directionpartly because of the general evolutionof our culture, but also because more automatedsystems seem to require it. As the level of automation increases, systems become more vulnerableto unforeseencontingencies, and the workers'role shifts fromproviding effort to monitoringthe systemand respondingto thesecontingencies (Hirschhorn, 1984).

The challenge to the commodityform of labor posed by this socializationlies in the fact that this quality,the importanceof which seems to be growing with automation,can have no economicvalue determination:the "conditionsof production"of a "sense of responsibility" are radically indeterminate,and measurementof the quantityof skill via trainingtime () becomes impossible.

Now, the underlyingefficacy of the wage-formis based on a genuine~if somewhatelementary- concept of justice: wage levels should reward investmentsin human capital. But when critical skills cease being behavioraland become instead attitudinal,the wage-determinationyardstick shifts from trainingtime and the anonymousrealm of to the more subjectiverealm of 30 BERKELEY JOURNAL OF SOCIOLOGY politics and ethics. With this shift,an importantelement of the legitimacy of the wage system-its anonymityand resultant objectivity-is undermined.

Beyond this destructivemovement, this socializationof work might also explain what could prove to be a sustained trend toward higher levels of concern for corporate culture and consensus-based decision-making. Moreover, an associated hypothesis would suggest that the only possible basis for consensus and for the sustained, as opposed to occasional, cooperationit requires,is broaderworker participation in decision- makingprocesses.

(3) Abstractionof Expertise

The abstraction of work is driven by the fundamental peripherizingnature of automation:workers' tasks shift from fabricationto controland interfacefunctions. Their contribution shifts from manual/roteexecution of pre-specifiedroutines to problem-identificationand problem-solvingactivity. Among Cain and Treiman'sfactors, we can thusexpect the importanceof motor skills and physicaldemands to diminish(1981).

On the one hand, this takes us beyond what Marx (1963) called "craftidiocy"- the model of the craftsmanat the centerof the productiveact becomes obsolete,and is replacedby thatof the operator who "controls the controls" (Hirschhorn,1984). The associated increasesin complexityand responsibilityreinforce the trendswe have alreadynoted.

But on the otherhand, and more intriguingly,abstraction risks underminingthe "use-value"of work to the workers-its intrinsic, humanly satisfyingvalue. When employees are no longer the principal agents of production-instead,standing along-side it, often as mere "buttonpushers"-it is more difficultfor them to confirmany sense of self-worthby referringto theirown role in making the product.As numerouscase studies of control-room operators have shown, automation means that boredom can become a real problem even when other dimensionof the job evolve towards greater responsibilityand complexity. This boredomcan reasonablybe hypothesizedto have a negativeeffect on employeemotivation.

The problemis not so much in the non-tangiblenature of the tasks. Managerialand professionaltasks are not tangible,but they pose no intrinsicmotivational problem. That is because managers and professionalswork on the system,not merelyin and around it; the goals of theiractivity- designing and directingthe overall system-aretherefore quite concreteto them. But for the lower- level employees,since theirown dexterityor "art"is no longerat ADLER: AUTOMATION AND CAPITALISM 31 the centerof production,goals risk becoming very abstract,and theirrole risks appearingto them as "mere work." Interestingly, thereappears to be a growingnumber of jobs like controlroom operatorsthat are simultaneouslyboring (most of the time) and highlyresponsible and complex.

In renderingwork intrinsicallyless interestingfor (larger?) numbersof workers,automation might thus be hypothesizedas underminingthe work ethic itself, and thereby undermining anothercritical condition of viabilityof the commodityform of labor. Many low-endcomputerized jobs are simplyso boringas to make themimpossible to staffon a full-timebasis. The ergonomie problemsof such work, the difficultyof maintainingacceptable quality levels without tangible motivatingelements, and the scarcityof people willingto work such jobs on a full-timebasis, all suggest that automation(in conjunctionwith broader social trendsin work expectations)might be undermininganother key mechanismof the invisiblehand: the supplyof labor may not ever matchdemand within feasible wage rates.

Part of the remedyto thisproblem might be a higherdegree of association of workers with the overall objectives of the organizationto compensatefor the loss of theiridentification with their own specific tasks. We have already noted this possible element of the post-capitalistfirm in the form of worker participation.

But anotherpart of the remedy might have to be shorter workinghours at least for a significant-andperhaps growing- segmentof the .And this too, would subtractfrom the centralityof the market.

(4) Interdependence

Let us hypothesizethat automationand productivitypressures tend to drive work organizationfrom the stage of stand-alone independentwork stations,first to sequential interdependence- where workersdepend on otherworkers further up the assembly line-and beyond that, into a stage of systemic, reciprocal interdependence characteristicof many computer-integrated operations (Thompson, 1967). Beyond the increase in the importanceof interpersonalskills, this greater interdependence seems to call for a new definitionof individuality-and therefore fornew typesof incentives.

As systemicinterdependence develops, the division between intellectualand manual work,between conception and execution, becomes progressivelymore blurred:the manual execution tasks are infusedwith a highertechnical and managerialcoordination content,and technical staff has to cooperate with workers to 32 BERKELEY JOURNAL OF SOCIOLOGY develop and debug new products and processes. Managers are forced to emphasize the productiveaspect of their roles and deemphasizethe monitoringaspect. Workerscan thusno longerbe managed as passive, technicalinputs, and managementneeds to acknowledgethem as active stakeholdersin the firm.

With this hypothesizedtrend, problems might be expected to arise for the individualizedform of the wage- a foundationstone of the commodityform of labor (Polanyi, 1957). When workers' contributionsare interdependent,group incentives appear more attractive;but group bonus systemschallenge the individualized wage-form.Second, such recognitionputs the autonomyof the worker collective in a very powerful position vis-a-vis management.Third, to the extentthat team-work and othersocial skills become more important,interdependence poses the same difficultiesof objectivemeasurement as responsibility.

These trendssuggest the emergenceof a new, more socially dense sense of individuality,one thatexpresses a new definition and balance of the individual and the collective dimensionsof workeridentity and interests(Lodge, 1975; Unger,1975). 7. Conclusion

The preceding discussion has argued that skill has several distinct dimensions, and long-run automation trends seem to encouragedistinct shifts along each of these. These shiftsmay be supported or inhibited by other developments, especially developmentsat the broader societal level; but technological change in the capitalistlabor process seems to generatechanges in work that not only tendentiallyincrease skill levels, but also underminethe viability of the human capital, market-driven regulationof the value of work. Automationseems to undermine the commodityform of labor in severalmutually-reinforcing ways: complexityinduces greaterfirm-specificity which underminesthe effectivenessof the labor marketas a mechanismfor matching supply and demand; responsibilityundermines the legitimacyof the human capital yardstickof wage determination;boredom underminesthe efficacyof the wage as a critical force in the mobilization of labor; interdependence undermines the individualityof the wage-form.

Alongside these destructiveeffects, however, I have identified some trendsand pressureswhich seem to prefigurea new "post- capitalist" form of organization: the subordinationof market mechanismsto negotiatedethical, political, and technicalnorms as well as the emergenceof demandsfor increaseddemocracy in the workplace,for free time,for recognitionof workersas collective stakeholdersin the firm. ADLER: AUTOMATION AND CAPITALISM 33

My argumenthas been primarilya technologicaldeterminisi one. There is good reason to reject technologicaldeterminism in anythingbut the verylong run; but my time-frameis preciselythe secular process of change in fundamentalsocial forms.Even in this time-frame,technological determinismis subject to the criticismadvanced by Levine and Wright(1980) against Cohen's (1978) presentationof Marx's theory.Levin and Wrightshow that social change, and, in particular,changes of mode of production, are created by class activity,not by anonymous forces like "technology."They show furtherthat the capacityof classes to act in a concertedmanner is above all a political and ideological matter,not primarilyan economic/technicalone. This critique ignores, however, the role played by economic/technical developmentsin fashioningthe capabilitiesthat workersbring to the political task. Such a propositionentails no claim that the forcesshaping class capabilitiesfrom outside the labor process are unimportant,or even that they are less important.Nor does this propositionsuppose that previous mode-of-productiontransitions were similarlypropelled. But it does suggestthat some elementof "technological determinism"-notto the exclusion of other determinations-maybe a fruitfulway of approachingthe analysis of longerrun trends.

The dimensions of skill I have identifiedpermit us to understandthe manner in which automationis shaping class capabilitiesin the labor process, by increasingtechnical culture (complexity), by expanding responsibility(socialization), by broadeningintellectual horizons and by encouragingrecognition of labor per se as a historicalproblem (abstraction), and by forging new forms of group identitythat allow for higher forms of individuality(collectivization).

If these are indeed the generaltrends in the evoltionof work, then the critique of Bravermanto date has not been radical enough.Braverman's vision of the progressivedegradation of work leadingto revoltcan be contrastedwith the thesisthat Schumpeter (1976, p. 162), in what might be one of the most insightful exegeses of Marx, formulatedthus:

The capitalist process not only destroys its own institutional frameworkbut it also creates the conditions for another. Destructionmay not be the rightword afterall. Perhaps I should have spoken of transformation.The outcomeof the process is not simplya void that could be filled by whatevermight happen to turnup; thingsand souls are transformedin such a way as to become increasinglyamenable to the socialist formof life. With every peg from under the capitalist structurevanishes an impossibilityof the socialist plan. 34 BERKELEY JOURNAL OF SOCIOLOGY

References

Adler,P. 1981 Automatisationet Travail: Le Cas des Banques. These de doctoratde troisièmecycle, Universitéde Picardie,France.

1985 "Rethinkingthe Skill Requirementsof New Technologies,"in D. Whittington(ed.) High Hopes for High Tech. North Carolina: Universityof NorthCarolina Press. 1987 "Automationand Skill: New Directions,"International Journal of TechnologyManagement 2, 5-6. Braverman,H. 1974 Labor and Monopoly Capital. New York: Monthly Review Press.

Bright,J.R. 1958a Automation and Management. Boston: Harvard Graduate School of Business Administration.

1958b "Does Automation Raise Skill Requirements?" Harvard Business ReviewJuly-August. Cain, P.S. and D. Treiman 1981 "The 'Dictionary of Occupational Tides' as a Source of Occupational Data," AmericanSociological Review 46 (June), pp.253-278. Cohen, G.A. 1978 s Theory of History: A Defense. Princeton: PrincetonUniversity Press. Cossalter,C. and L. Hezard 1983 "Nouvelles Perspectivesde l'Informatisationdans les Banques et les Assurances,"Centre d etudes et de Recherches sur la Qualifications,Collection des Etudes,No.4.

DelegationNationale CFDT. 1985 Livre Blanc sur l'avenir du CreditLyonnais. Paris.

Doeringer,P. and M. Piore 1971 Internal Labor Markets and Manpower Analysis. Lexington, MA: D.C. Heath.

Durraffourg,J., F. Guerin,F. Jankovskyand B. Pavard 1979 "Analyze des Activitiesde Saisie Correctionde Données dans rindustriede la Presse,"Le Travail Humain 42:2.

Edwards,E. and F. Lees (eds.) 1974 The Human Operator in Process Control.London: Taylor and Francis.

Edwards,R. 1979 ContestedTerrain. New York: Basic Books. ADLER: AUTOMATION AND CAPITALISM 35

Fiske, E.B. 1982 "ComputersAlter Lives of Pupils and Teachers," New York Times April 4. Form,W. 1980 "Resolving Ideological Issues in the Division of Labor," in H.M. Blalock, Jr. (ed.) Sociological Theory and Research: A CriticalAppraisal. New York: Free Press. Galle, R. and F. Vatin 1980 Le Modele de Fluidité: Etude Economie et Sociale d'une Raffineriede Pétrole. Bandoli Laboratoriede Conjonctureet de Prospective. Gordon,R., R. Edwards,and M. Reich 1981 Segmented Work, Divided Workers. New York: Cambridge UniversityPress.

Grandjean,E. and E. Vigliani (eds.) 1980 Ergonomie Aspects of Visual Display Terminals. London: Taylor and Francis. Hirschhorn,L. 1981 "The Post-industrialLabor Process,"New Political Science Fall.

1984 BeyondMechanization. Cambridge: M.I.T. Press.

Knights,D., H. Willmott,and D. Collinson (eds.) 1985 Job Redesign: Critical Perspectives on the Labour Process. London: Grover.

Lawler, E.E. m 1981 Pay and OrganizationDevelopment. Reading: Addison-Wesley.

Levine, A. and E.O. Wright 1980 "Rationality and Class Struggle," New Left Review 123 (September-October).

Lodge, G.C. 1975 The New AmericanIdeology. New York: Knopf. Marx, K. 1867 Das Kapital 1st edition.Hamburg: Verlag von Otto Meisner.

1963 The Povertyof Philosophy.Moscow: InternationalPublishers.

1973 Grundrisse,transi. M. Nicholas. Harmondsworth: Penguin.

1981 Capital, vol. 1. New York: Vintage. Matteis,R.J. 1979 "The New Back Office Focuses on Customer Service," Harvard Business Review March-April. 36 BERKELEY JOURNAL OF SOCIOLOGY

Noble, D. 1984 Forces of Production.New York: Knopf. O'Brien, J. 1968 The Impact of Computerson Banking.Boston: Bankers.

Polanyi,K. 1957 The Great Transformation.Boston: Beacon.

Schumpeter,J. 1976 Capitalism, and Democracy.New York: Harper. Shaiken,H. 1985 WorkTransformed. New York: Holt, Rinehartand Winston. Shiel, B.A. 1981 MCopingwith Complexity" Palo Alto Research Center,Xerox.

Spenner,K. 1983 "DecipheringPrometheus: Temporal Changes in the Skill Level of Work," American Sociological Review 48 (December), pp.824-837. Steiber,J. 1959 The Steel Industry Wage Structure. Cambridge: Harvard UniversityPress.

Thompson,J.D. 1967 Organizationsin Action.New York: McGraw-Hill. Treiman,D. and H. Hartmann 1981 Women, Work, and Wages. Washington, D.C.: National Academy Press.

Unger,R.M. 1975 Knowledgeand Politics. New York: Free Press. Weiner,E. 1985 "Beyondthe SterileCockpit," Human Factors February. Wood, S. (ed.) 1982 The Degradation of Work?London: Hutchinson. Yellen, J.L. 1984 "Efficiency Wage Models of ,"American Economic Review 74, pp.200-205. Zuboff,S. 1982 "New Worlds of Computer-mediatedWork," Harvard Business Review September-October.