3.1.1 Long-term compensation for initial job losses

Concerning possible job losses caused by digitization research refers basically to the well-known debate about technological unemployment. As John Maynard Keynes put it the increase of technical efficiency takes place faster than we can deal with the problem of labor absorption (Keynes 1963). In regard to this question, there is the often-cited “conventional wisdom” of labor-market research going back to Ricardo, according to which short-term negative employment effects of technological change are always compensated for in the longer term by efficiency gains, new products, new markets and new employment opportunities (e.g. Agion and Howitt 1994). Thus Evangelista et al. (2014) see, on the basis of a detailed literature review in anticipation of the adoption of digital technologies, little clear impact on employment. In particular, they emphasize that it is particularly difficult to attribute causal effects on employment to this technology. The reason for this lies in their potentially widespread use in many areas. Although substantial empirical studies are lacking, in the literature predominates an optimistic view of the long-term employment effects of digital technologies. This is because one must see both the directly negative and the indirectly positive effects on employment as a result of efficiency improvements and price reductions, and the opening up of new markets (Evangelista et al. 2014, p. 806).

A similarly positive view of the impact on employment can be found in the context of Germany’s 4.0 debate. Here are predicted not only generally high productivity gains and economic growth rates (e.g. Bauer et al. 2015), but with that also consistently positive employment effects. Thus, on the results of the study by Spath et al. (2013, pp. 46) the vast majority of industrial enterprises assume that human labor will remain significant in industrial production in the next few years and not be reduced. The same is found in a study by the Boston Consulting Group: “In our analysis of Industry 4.0’s impact on German manufacturing, we found that the growth it stimulates will lead to a 6 % increase in employment during the next 10 years…. And demand for employees in the mechanical engineering sector may rise even more—by as much as 10 % during the same period” (BCG 2015, p. 8). Following this study, job growth in German manufacturing between 2015 and 2025 will amount numerically to about 390,000 jobs.

3.1.2 Sweeping job losses

In contrast, other authors argue that this employment compensation mechanism is no longer at work in the labor market in conditions of today’s diffusion of digital technologies, as it was in the past. Thus Brynjolfsson and McAfee (2014, p. 177 ff.) emphasize that the increasingly rapid development and diffusion of digital technologies implies an increasingly widening gap between the new demands of technology and the generally more slowly effective socioeconomic adaptation mechanisms and the implicated opportunities for employees and institutions. The authors speak of an “…inability of our skills, organizations, and institutions to keep pace with technical change” (ibid.: 178). In addition, however, also the volume of available jobs for job seekers is dwindling increasingly, since the demand for many tasks and qualifications is dropping. Many jobs are being replaced by digitization, creating increasing job losses in those areas. The authors see their assumptions confirmed by the rising labor productivity and falling employment figures from the first half of the last decade (ibid.: 165).

Frey and Osborne reinforce this view with an analysis of the US labor market, which shows that very significant potentials for job losses go hand in hand with digital technologies. With regard to recent advances in “machine learning” and “mobile robotics”, the authors have developed a methodology to categorise occupations according to their susceptibility to computerization. This methodology is used to assess the probability of computerisation for 702 occupations, and the impacts of computerisation on US labour market outcomes. The authors differentiate between high, medium and low risk occupations, depending on their probability of computerisation. Their argument is that not only simple, but in particular routinizable tasks including more complex activities can be automated by the new technologies. Therefore, they conclude that almost half of all current jobs from the most various economic sectors could be substituted in this way. Their central message is that approximately 47 % of all activities on the American labor market over the next one or two decades are potentially threatened by automation (Frey and Osborne 2013, pp. 38.). Frey and Osborne speak of two different waves of far-reaching effects of computer applications on employment in the next few decades: the first wave comprises a pattern of progressive automation, namely the substitution of primarily routinizable and at least partially non-routinized activities by digitized technologies in the most various sectors. They expect a second wave of automation to follow, which will spread to activities comprising creative and socially interactive tasks. Their expectation is that, in contrast, “…most management, business, and finance occupations, which are intensive in generalist tasks requiring social intelligence, will be largely confined to the low-risk category. The same is true of most occupations in education, healthcare, as well as arts and media jobs.” (Frey and Osborne 2013, p. 40)

On the basis of the same concept analysis Bowles (2014) calculates similarly high job substitution risks for the European labor market. He differentiates between different country groups in Europe. With regard to the German labor market he concludes that more than 51 % of all activities were at risk of being replaced in the long term by automated processes through digitization. The reason for this high percentage is the importance of industrial work that may be threatened in the near future by especially rapid digitization spurts. A study by INGDI Bank on the prospects for the German labor market comes to yet farther-reaching assumptions (ING DiBa 2015). This study, oriented on the analysis concept of Frey and Osborne, “investigated approximately 81 % of low-level” and marginally employed persons in Germany. On this basis the study arrives at the statement that about 18.3 million, or 59 % of these jobs are at risk in their current form from the progressive technologizing of the German economy. The study indicates these are most often transport and storage activities, general auxiliary personnel, office and service-sector work. The authors of this study emphasize however at the same time that this is not expected to lead to abrupt job losses; rather, they expect a process of slow substitution or exchange, as new technologies, in particular easily deployable robots, come to dominate only gradually. The principle however is to assume a fundamental transformation of the labor market as a result of digitization and Industry 4.0.

3.1.3 Relativizations

It is not surprising that the outlined forecasts of far-reaching job losses through the application of digital technologies in the political and scientific debate are controversial to the extreme. Although some authors in the context of the “Industry 4.0” debate also assume that the employment volume in total could shrink, they reject the volumes of job losses mentioned. The critical arguments listed relate primarily to the fact that in these labor market forecasts it was purely a potential assessment of technological development and static analyses of occupations; ultimately, these are “automation probabilities” for existing occupations, without regard to the specific operating conditions involved. Above all however, some critics say studies such as those of Frey and Osborne (2014) neglect the peculiarities of the actual activities of the different occupational groups which, in an assessment of their activities substitutability, can lead to differentiated results. Thus, starting from these considerations Bonin et al. come to significantly lower predictions: the share of jobs with a high level of automation probability in the US is at just 9 %, while in Germany this is still only 12 % (Bonin et al. 2015, pp. 14). In a recently published studies by Dengler and Matthes present similar findings for the German labor market on the base of a further developed methodology. They calculate the substitution potentials of occupations based on German occupational data from an expert data base of the Federal Employment Agency. Following their results 15 % of employees have a high substitution potential in the year 2013 in Germany; these employees are employed in an occupation in which more than 70 % of the tasks could be substituted by computers (Dengler and Matthes2015).²

In the opinion of other critics, the assessments of possible job losses above all completely neglect the highly qualitative job skills required for working with complex technologies. Their argument is that because these are indispensable to the efficiency of digitized processes, they present an obstacle to far-reaching automation attempts. This is pointed out particularly by Sabine Pfeiffer and Anne Suphan (2015), who show on the basis of their “work capacity index” that the greater part of the workforce is able to deal with technology related complexity demands on the basis of available experiential resp. implicit knowledge, and that such workers are therefore crucial to coping with the challenge of digitization. They name here a share of 71 % of all employed persons in Germany (ibid.: p. 222). The conclusion is that this activity share is relatively stable and secure against the presumed far-reaching substitution effects. Beyond these concrete numerical data that relativize far-reaching effects, there are more fundamental arguments that put into question extensive substitution (see also part 4.1). It is pointed out that technological potentials are often overestimated, that both professional as well as activity structures have a longer-term high dynamic that works against linear automation effects, and finally, that macroeconomic adjustment processes, as also emphasized by the representatives of the compensation thesis (see above), are overlooked.

3.2.1 Upgrading of qualifications

The one pole describes a development path by which the digitization of work generally brings an appreciation, or an “upgrading” of required worker qualifications. This development perspective is represented relatively strongly in both the scientific as well as public debate. Following the study by Zuboff (1988, pp. 10) on the consequences of the use of information technologies, an upgrading of qualifications is seen to take place in two ways:

On the one hand this is considered the result of progressive computer technology automation of simple jobs that are extensively substituted. The prerequisite for this is, that it be routinized and to a large extent highly rule-based work that therefore can be taken over relatively easily by computer algorithms. As aforementioned, Frey and Osborne (2013) speak in this context of waves of extensive substitution effects of computer use that are foreseeable for the coming years particularly in these activities. This upgrading model is referred to in the literature also as “skill-biased technical change”—i.e. the winners in the progressive substitution by digitized technologies are those groups of workers who already feature higher qualifications and more behavioral resources (see also Brynjolfsson and McAfee 2014, p. 136).

On the other hand, upgrading can be understood as a process that captures all employee groups in general. Digitization of work in this perspective is a process of the computerization of work which makes increasingly available a wide variety of information about ongoing processes. Their complexity and possible uses result fundamentally in new and hitherto unknown requirements for all activities. Zuboff therefore speaks of the growing importance of “intellective skills” based above all on a theoretical understanding of processes, i.e. the ability to grasp the prerequisites and consequences of the use of data available at any time (Zuboff 1988, p. 94). The general consequence would be “better jobs—jobs that at every level would be enriched by an informating technology” (ibid.: 159). Zuboff describes this development perspective that information technology has the unique capacity to “informate” nearly all activities and jobs (ibid.: 10).

These tendencies towards qualification upgrading are reflected in the literature, especially for current information technology applications of the “Internet of things”, because such systems provide, through their data acquisition and evaluation, a degree of transparency over production processes in a previously unknown way (Zammuto et al. 2007; Evangelista et al. 2014; Boos et al. 2013). Thus also in the public and innovation policy debate over Industry 4.0, it is also emphasized that in future, a general upgrading of qualifications will be possible and also certainly will occur. Here reference may be made to Henning Kagermann as representative of a great variety of authors and positions, and one of the leading advocates of Industry 4.0 in Germany, in whose opinion people in the future will be employed less as “machine operators” than as “more in the role of mediators of experience, as decision-makers and coordinators…[that is,] the variety of job content for the individual employee will increase” (Kagermann 2014, p. 608; e.g. also Bauernhansel 2014; Wissenschaftlicher Beirat 2014).

This perspective on the development of work corresponds to a model of work organization at the enterprise level characterized by a high degree of structural openness, a very limited division of labor and high flexibility. For, as labor sociological findings show, this is the organizational precondition for the qualifications and experience of employees to be brought to bear in the ongoing system and, above all, for them to be able to cope at any time with unanticipated incidents and particular situations through competent and experience-tested action. In the debate over the digitization of work, one model of work organization is variously emphasized for its perspective on the evolution of change in work: holistic organization or, metaphorically, also swarm organization (Hirsch-Kreinsen 2014). This organizational model is characterized by a loose network of qualified and equally entitled employees. Simple and low-skill jobs are not to be found because they have been (or will be) largely replaced by automated processes. The central feature of this organizational model is that there are no defined tasks for individual employees. Rather, the work “collective” functions in a self-organized, highly flexible and situationally determined way, varying its behavior with the problems to be solved in and on the technological system. However a frame of action exists, predetermined by top-level management, prescribing the basic rules of action, strategic objectives, collective orientations and guiding policies, for an optimally trouble-free and smooth technological process (Neef and Burmeister 2005).

In other words, this model of work organization aims for the explicit use of informal social processes of communication and cooperation and the associated extra-functional skills and the accumulated specific process knowledge of employees. To follow Böhle (1992), this organization model is based on the necessary interplay of general social capabilities such as communication and methodological skills with practical work-process knowledge specific to the given manufacturing process.

3.2.2 Polarization of qualifications

The other pole is represented by the notion of the “polarization” of job activities and skills. Its central developmental mechanism is an increasing erosion and replacement of medium-level skills. Accompanying this is a growing share of demanding, high-qualification activities on the one hand, and the preservation and partly the expansion of easier but not routinized and therefore not automatable activities on the other. This thesis is particularly prominently represented by the author, as well as Dorn (2013), Collins (2014) and Brynjolfsson and McAfee (2014) who, especially in macro-structural perspective point to the development of the US labor market, but also variously to transformations in the EU labor market (e.g. Goos et al. 2009; Bowles 2014). On this view, generally in the last two decades the proportions of demanding jobs in managerial, technical and professional occupations, but also those of less demanding activities in the service and industrial sector, have increased, while the shares of medium-skill job areas such as sales, administration and industrial production have stagnated or decreased.

The reason for this development is that not only simple, routinized activities, but especially also many activities on intermediate skill levels can be automated and thus substituted. The prerequisite however for this is that these should be activities of a well-structured and rule-oriented character and can therefore be suitably algorithmized. Specifically, these have been usually quite demanding types of production work such as machine, e.g. automotive installation and system monitoring, as well as many routine administrative and service activities on medium skill levels (Autor 2010; Marin 2014; for an overview see Autor 2013). As already described in the context of the consequences of the new technologies for the labor-market (see Part 3.1), through the new technologies also relatively routinized work in segments of mid-range complexity and income is also becoming increasingly automated. Complex activities in high-wage areas such as management, consulting or financial services, and low-wage jobs such as simple manipulation and monitoring activities, nurses and geriatric nurses, however, enjoy high demand (Marin 2014). Goos and Manning characterize this trend as the emergence of “Lousy and Lovely Jobs” (Goos and Manning 2007).

This macrostructural justification for a polarizing trend is mirrored by research that focusses on the workplace level, such as a broad German study on the development of skilled-worker jobs in industrial companies, whose authors fear a progressive “de-skilling and part-substitution” of intermediate skill levels as a result of information technology automation in those areas. They can speak at best only of the uncertain destiny of such “residual categories” of skilled labor which are activities that cannot be automated or would cause a disproportionate expense in order to be automated (TA 2007; Kinkel et al. 2008; Düll 2013). Similarly argue Windelband et al. (2011) in their study of work in the context of intelligence-networked logistics systems, which creates at the enterprise level the tendency towards a differentiated structure of activities between sophisticated, high-skill jobs on the one hand, and some remaining but devalued specialist tasks and non-automated simple activities on the other hand. This deskilling is also referred to as “Digital Taylorism” since the digital technologies allow an optimization of Taylor’s principles of work simplification and work control especially for complex tasks (The Economist 2015, p. 63). However, also new jobs for low-skilled workers may arise thereby. It can therefore be assumed that low-skilled employees will be not, as the Upgrading thesis assumed, largely substituted by digital technologies.³

This perspective of polarization corresponds at the enterprise level to a work-organizational design model with a strongly exhibited division of labor. On the one hand it is characterized by a low number of simple activities with little or no room for maneuver, such as standardized monitoring and control tasks. On the other hand, an extended or even newly created group of highly qualified experts and technical specialists emerge whose skill level is well above the previous skilled-worker level. These workers are responsible not only for dispositive issues such as coping with disturbances, but they also take on various tasks of production management. This model of work organization corresponds largely to currently prevailing forms of work in many high-tech companies, which can be characterized as a contradictory combination of the design principles of decentralization and task-area broadening on the one hand, and of structuring and standardization on the other hand (e.g. Kinkel et al. 2008). In this way companies avoid high-risk or uncertain organizational innovations by following this already established path of work organization design. This model can thus be called in abbreviated form polarized organization.

3.2.3 Developmental alternatives

To summarize, it can be said that some very divergent development perspectives for digitized work can be argued. Quite obviously it is not possible to speak of the “one best way” of task and organizational design. It should also be stressed that the outlined poles of the spectrum of possible trends and work organization models denote conceivably extreme cases in the future situation. Rather, it is to be assumed that, in dependence on specific application conditions, system functions and structural workplace conditions, diverging and intermediate patterns of work organization will attune themselves to one another and ultimately become the object of company decision-making processes.

However, in labor research, plausible reasons for the possible relevance of swarm organization can be given in the context of the adoption of digitized production systems: An effective system control by skilled workers can be assured with minimally regulated, informal and cooperative forms of work processes (Lee and Seppelt 2009; Cummings and Bruni 2009). Moreover, it is quite possible in the context of such a model of work organization to maintain control over real-time decision-making and communication processes (Spath et al. 2013, p. 115). In addition it is emphasized that complex systems change states “spontaneously” and have intransparent and unpredictable effects (Grote 2005) that require highly flexible workplace interventions that are as unplannable as they are uncontrollable. In addition, it is assumed that the lengthy introduction and start-up phases of Industry 4.0 systems are due to their complexity, in the course of which activities and work organization must show a high degree of flexibility and problem-solving skills, while still far from able to reach a definable (end-) state. Finally, the “lifecycle” of complex systems is referred to, which can always involve new system states that are difficult to control; both unexpected start-up problems, as well as current problems and unexpected disturbances in normal operation can subsequently only be overcome in the context of open and informally designed forms of work (BMWi 2013).

Finally, in this context, it should be asked what effects the information-technologically possible temporal and spatial separation of work functions from the real process, and above all the possibilities of their temporal and spatial flexibilization will have for the design of work and the distribution of capabilities (Kinkel et al. 2008, p. 245). This option applies to the activities at the shop-floor level as well as those at higher hierarchical levels, because the boundaries of organizational structures are confused in this way; it is probably becoming more and more difficult to speak of well-defined models of work organization and company hierarchy, while work processes are taking on an increasingly informal and unstructured character. Communication and social interaction in the work process will perhaps then be mediated mainly or only through IT and other media-applications, and the above model of swarm organization will possibly predominate as the leading form of largely unbounded work. Thus new forms of cross-company distributed, Internet-based, and in tendency global labor relationships are conceivable, which have recently been discussed under the heading of “crowdwork” (e.g. Leimeister and Zogaj 2013; Benner 2014).