In examining the literature on the impact of computers on organizations, we find it puzzling that many people are willing to speak and write as though the overall effects of computing technologies were a foregone conclusion. Many observers seem to believe that computer impacts can be determinedapriori by deducing fromabstractprinciples whatthe effects of computersareboundtobe. Wearguetheopposite: thatevidenceonthesesubjectsisactually fragmentary and very mixed, and that a priori arguments are particularly inappropriate in light of the wide range and variety of variables at work in these situations. In the following pages we examine the literature on the effects of computing on the numbers and quality of jobs on management decision-malcing and on organizational dealings with clients and customers. We alsoconsidervariousperspectivesonthecausesoforganizationaldecisionsto adopt computing in the first place. Our conlusions are similar for all of these areas: virtually none of the studies mounted to date has been capable of yielding a persuasive and comprehensive view of computer-induced social change. We need to go beyond individual case studies, to initiate a program of comparative research on representative samples of organizations. QUALITY OF WORK The research literauire on the impact of new information technologies onjobcontentandjobsatisfactionprovides amass ofcontradictoryfindings.The wide range of informed opinion can best be illustrated by describing the two extreme positions: deskilling versus upgrading. The deskilling perspective suggests that automation is used to striprelatively-skilledjobsoftheirconceptualcontent(Braverman).Thoseconceptualtasks previously integrated into work are either built into computer algorithms, or are transferred to a numerically smaller number of high-level specialists. This deskilling manifests itself in two distinct ways: intro-occupational changes, where the skill content of a particular job decreases overtime; andinter-occupational changes where the numberof personsin skilled jobs shrink andthenumberof employees inunskilled jobs rises. Inthe secondofthese cases, one empirical indicator of computer-generated deskilling is a shift in the occupational structure of the white-collar workforce. The deslcilling position implies a more polarized pyramidal distributionof skill: a mass of unskilled clerical workers atthe bottomand a small number of conceptual workers at the top (Driscoll). Kraft and Greenbaum have taken the analysis even further, arguing that even conceptual jobs like programming are being increasingly deskilled. In contrast, Guiliano andothers have arguedthat computerization and othernew information technologies upgrade rather than deslal white-collar workers The upgrading thesis suggests that automation primarily occurs in already-routinized contexts; the new technology takes the drudge work out of information processing by automating the most repetitious manual aspects, leaving humans to concentrate on conceptual and decision-making tasks. The potential victims of such upgrading are the lowest levels of clerical workers who manually manipulate data However, this lowest stratum need not be adversely affected by automation because the introduction of computers manifests itself in the relative growth of higher-level jobs and the relative shrinkage of lowerpositions. The absolute number of low-level workers need not decline, because total white- collar employment is still growing. Thus, in direct contrast to the deskilling approach, the impact of computer technology is said to be an

increase in worker satisfaction, and a shift in occupational structure away from a pyramid shape (few skilled, many semi- or unskilled) toward a diamond shape (few top managers, many professionals and middle managers, few clericals) (Zuboff). Many case studies of intra-occupational change describe loss of conceptual content, fragmentation, and deskilling of white-collar work after computers are introduced. However, some observers also find consolidation of tasks, and upgrading. Attewell's study of the insurance industry confirms that both upgrading and deskilling occur within occupations being computerized, but finds that upgrading predominates Several quantitative studies of the whole economy find little evidence for intra-occupational desldlling since 1949, but these studies are based on government data whose quality has been disputed Evidence on interoccupational change is similarly contradictory, althoughmostcase studiesreportupgrading. Quantitative studies based on disputed DOT data don't find deskilling. Studies which ask workers about their experiences with computerized work are typically positive, although some report increased time pressure and increased supervision. We conclude that both deskillingand upgrading canoccurfollowingcomputerization. Theoretically, whatmattersis to find out what factors and situations produce these vatious outcomes, and the relative frequency of each effect We propose systematic surveys to answer these questions. EFFECTS ON EMPLOYMENT Pessimists anticipate substantial unemployment due to the labor saving effects of new information technologies. Several input/output models of European economies confirm this, albeit based upon shaky data Certain case studies show employment losses of up to 50% in fields like metalworking. But these studies are not from representativesamplesof businessesand somustbetreated with caution Optimistspointto earlier periods of technological change when rapid growth in productivity did not create unemployment If goods become cheaper, demand grows, and total production increases Thus, even with more pivductive technology one still needs as many workers as previously. This optimistic position uses the concept of"long waves" of economic boom and bust, in which the introduction of new technologies (steam power, electricity, automobiles, microelectronics) causes sudden surges in investment, an upswing in economic activity, and increases in employment At present, insufficient evidence exists to decide between the optimistic and pessimistic analyses. We need careful studies of representative samples of firms, documenting their employment levels at various stages of computer automation, to properly evaluate the employment impact of the new technologies. MANAGEMENT EFFECTS Students of organizations have frequently observed that control over information is a source of powerin organizations. As such, new technologies that alter the quality and availability of information are likely to shift balances of power between organizational groups: workers, supervisors, middle managers executives (Olson and Lucas). Laudon and others viewed such processes as leading to increased centrolization of power in computer-automated organizations. Leavitt and Whisler predicted that new information technologies would eliminate whole layers of middle management as improved information led to centralized decision making higher up the corporate heirarchy. Several case studies have supported this view; and recent research on MIS and computer mail emphasize"top down" controL However, there is some evidence for the opposite view, that the increase in communication resulting from new technologies can decentralize decision making. Blau et aL have found that, far from eliminating levels of management computers are associated with an increased number of levels of line management and with enhanced local management decision making. In several studies of the introduction of computers in local governments Danziger, Dutton, Kling, Kraemer, and Northrop have documented subtle shifts in power among supervisors, bureaucrats, andmanagers. Contextualvariables werealso foundtobeimportant the effects in small municipalities were not the same as those in larger ones, for example. Robey offers a complementaly view, arguing that sometimes computers don't effect the distribution of power at all sometimes they reinforce the status quo, sometimes they aid decentralization.

If we assume that there must be a single effect of computers on management then these case studies appear contradictory. However, if we assume a range of effects is possible, then the taskof future researchbecomes clear. We mustidentifythose variables whichcanaccount for differential outcomes through comparative research on representative samples of organizations, examining factors such as size, industry type, degree of prior routinization of work, skill-level of workforce, and soon ORGANIZATIONS AND THEIR PUBLICS Changing information flows will also alter relationships between organizations and their environments- particularly the general public. Rule, Shills, and others have focused on how computers whet the appetite of organizations, especially government:, for information on the people with whom they deal Others have speculated on new kinds of informational services which may become available to the populace because of computers. Much additional research is needed in this area. THE LMPETUS FOR INNOVATION There is a common belief that organizations adopt computer and information technologies in order to pursue goals of efficiency and costeffectiveness. Againstthis view is a position firstarticulated byE ulthattechnologyisa selfsustaining force, which generates needs for itself: once a technology is available it is inevitably used, applications are found for it Odd as this latter position sounds, there is certain evidence for it Several writers have found that whateverthe intentions of managers, computers do not necessarily save money or increase efficiency. Rather than improve activities already in place, computers may be adopted for, or give rise to qualitatively new organizational activities. The application and spread of computers can reflect power politics within organizations rather than efficiency and so on. Such preliminary insights also suggest the need for futire comparative research into the causes of the introduction of new computer-based technologies.