Abstract
Digitalization has two very different effects on work. On the one hand, it leads to a re-Taylorization of work, de-qualification and a loss of workers autonomy. On the other hand, digitalization of work leads to new forms of indirect control and algorithmic control that can be used to manage and instrumentalize the supposed autonomy of workers to actually enable an unequal and exploitative labour process. This article discusses the questions of heteronomy related to the digitalization of work, presents central aspects of new forms of control (direct, indirect, and algorithmic) and explains why formalization, data centred decision making and flexible structures are used to control the labour process and improve heteronomy of work.
Introduction
Digitalization has two very different effects on work. On the one hand, it leads to a re-Taylorization of work, de-qualification and a loss of workers autonomy. On the other hand, digitalization of work leads to new forms of indirect control and algorithmic control that can be used to manage and instrumentalize the supposed autonomy of workers to actually enable an unequal and exploitative labour process.
The first section outlines Marx’s theoretical framework which can be used to understand the impact of technological change on workers autonomy and the labour process. This framework is further developed in Labour Process Theory (LPT) (Braverman 1974; Knights & Willmott 2016; Smith & Thompson 2010). The main idea is that concrete labour must be controlled to produce surplus value but at the same time the labour process necessitates workers autonomy to a certain degree. Working contracts, as will be argued in the third section, reflect this problematic dialectic.
The second section outlines the concept of autonomy (theoretical and practical) and control (direct, indirect and algorithmic). Heteronomy is the antipode to autonomy and means that our beliefs, intentions or actions are controlled by someone else. Digitalization as a formalization allows management to restructure the work process and the use of data to automate control. This involves the expansion of control of the labour process and therefore increases heteronomy. The fourth section examines how formalization, data centred decision making and flexible structures are used to control the labour process. Data centred decision making and flexible structures use the felt autonomy of workers for controlling the labour process.
The final section then demonstrates the German Reference Architectural Model Industrie 4.0 (RAMI 4.0). This Model standardizes and fixates some of the concepts that are responsible for the loss of autonomy in the process of digitalization. As the German standard for digitalization projects within industry it can be understood as an example how capitalist objectives like hierarchies, value stream and life cycle are integrated into digitalization projects and become material structures for heteronomy within digitalization.
Work in Marx
The classical problem of transforming labour power into concrete labour was first theorized by Marx (1849, 1867) in order to explain how capitalists capture surplus value for profit. Today, researchers in Marxist labour process theory discuss – among other things – managerial strategies to control labour and optimize this transformation. The role of technology in this process is emphasized in LPT, and the loss of real autonomy for workers is problematized.
Digitalization is being put forward as a strategy to find a current solution to the transformation problem: transforming labouring capacity (Pfeiffer 2014) into labour power and labour power into concrete labour. Digitalization is an attempt to create relative surplus value. Therefore, a real subsumption of the labour process under capital is necessary (MEGA II/4.1). Marx indicated that contrary to formal subsumption, the labour process is changed in real subsumption. This real subsumption by digitalization does not only lead to a different form of concrete labour but also changes labour power and labouring capacity. The concept of labouring capacity can be seen as a helpful addition to the analysis of the labour process in the tradition of Braverman (1974). Workers resistance and autonomy in the changing labour processes under digitalization are closely related to the subjective qualities of labouring capacity.
Marx distinguishes between the use value of labour and its exchange value. While labour power is looked at from an exchange value perspective, labouring capacity is looked at from a use value perspective. Labouring capacity represents the subjective side of labour and labour power describes the abstracted objectified side of labour. The concept of labour as the production of use values by concrete labour is narrowed in capitalism because labour also needs to produce exchange value. And the production of exchange value is what defines labour which is subsumed under capital (MEW 23: 532). Not every labouring capacity and especially not every aspect of labouring capacity is part of labour power. This is the main reason why some forms of autonomy are even seen in alienated labour (Pfeiffer 2014).
In Marx’s early works Economic and Philosophic Manuscripts (1844, MEW 40 I: 465–588), labour always seems to be alienated labour, and/or labour which is subsumed under capital. But this is mainly a matter of terminology since even in the Economic and Philosophic Manuscripts, Marx describes life activity as something that could also be described as non-alienated labour. In Capital I (MEW 23: 192), he actually describes labour as non-specific to capitalism. Negt and Kluge (1981) and Pfeiffer (2004) see labouring capacity as historically specific (see also Marx MEW 40: 541f and 546) but not necessarily as alienated. Even when labouring capacity is transformed into labour power and the labour process is subsumed under capital, aspects of autonomy remain in the labouring capacity of the workers. This includes for example emotions, experiences, intuitions and creativity (Pfeiffer 2014). In his Mill excerpts, Marx sees this kind individuality and sensuality mainly in non-alienated labour (MEW 40 I: 462f). Jaeggi (2005) develops a normative and critical theory of society with the concept of alienation at its core.
Digitalization can be seen as a threat to the residuum of autonomy in labouring capacity and the concrete capitalistic labour process. This is due to a scientific management of areas that could formerly not be standardized, measured and controlled like knowledge work, complex collaborative and communicative work, affective labour and other. Scientific management’s inability to measure output or to attribute the result of collaboration to groups with highly interrelated work or not fully defined output results is an ongoing problem that whole areas like performance management need to deal with. Digitalization of work pretends to offer new solutions to this problem.
In Economic and Philosophic Manuscripts, Marx already develops the idea that a rising productivity of labour leads to a loss of autonomy because machines become more important and the worker more dependent on these machines (MEW 40: 511f). Pfeiffer (2014) argues,
That the organic composition of capital leads to a relative increase in dead labour not only has quantitative economic consequences, but also subject related and socially qualitative ones: It yields an increasing necessity of appropriative activities (i.e. expenditure/formation of labouring capacity) for mastery in dealing with more complex, accumulated dead labour, because an increment in accumulated, objectified labour no longer is just an increment of machinery, but even more an increment in complexity and abstraction levels – a process to which digitisation has made and will continue to make a decisive contribution. (p. 611)
So, for Pfeiffer labouring capacity, subjective elements of the workers and autonomy play even a bigger role in the context of digitalization. This might be a too positive analysis since the tendencies Marx describes are still present, de-qualification is common even in formerly high-skilled jobs and the subjective elements which are to a higher degree incorporated are also deformed by subsumption under capital (for this last aspect for example, Boltanski & Chiapello 1999).
Autonomy and control in digitalization
Baumann (2000) argues that a definition of autonomy is not universally agreed upon, but involves rationality and freedom which can be tracked back to Kant. Self-governance or self-determination is often used interchangeable with autonomy. Baumann distinguishes theoretical and practical autonomy. Theoretical autonomy includes epistemic autonomy and the practical side includes autonomy of the will and autonomy of action. These three aspects of autonomy also play an important role in the context of a digitalized economy.
Data and knowledge that are used to make economic decisions are usually only in the hands of management. Workers have normally only access to the data and knowledge that is necessary for their jobs although the new European General Data Protection Regulation would allow them access to a wider range of work related data. In the case of algorithmic decision making and the monopolization of data by external enterprises workers are expropriated of their data and knowledge to an even higher degree. This was already the case in scientific management, was reduced in group-work and humanization of work, increased (partly) with toyotism and is on the rise again with digitalization. Epistemic autonomy is therefore hard to achieve, if economic knowledge becomes more complex and the means (like data) to achieve it cannot be accessed because it is controlled by management or external enterprise. Workers have to trust or obey the management and algorithms. Furthermore, by setting standards and norms, digitalization also defines what there is to be known. It becomes harder to think ‘outside the box’.
Practical autonomy is also challenged by digitalization. Autonomy of the will is a complicated matter in work relations since it is never easy to say when it is lost. Turning labouring capacity into labour power and concrete labour is also a transformation of the will. The primitive accumulation of capital as discussed by Marx (MEW 23: 741–791) is already a case in which workers had to be formed and motivated. Many management strategies are deployed in order to create a working will (Boltanski & Chiapello 1999). Introducing democratic elements in enterprises (Sattelberger et al. 2015) is part of this strategy. Current methods even try to manage and control emotions of employees (Moore & Piwek 2017). Big data and data about individual employees allow to create worker profiles and statistical strategies of how to manage and nudge them. Gamification and a competitive work environment are supposed to motivate workers. Digitalization and subjectification of work lead to a deeper intrusion into the will of workers.
Autonomy of action depends heavily on the kind of work people are doing and the used management strategies. Highly qualified work necessitated a high degree of autonomy while low qualified work involves only a very small amount of autonomy. Digitalization is assumed to polarize jobs (Goos et al. 2016), so that there will be mostly high or low qualified work and almost nothing in between. Digitalization can involve different management strategies but at least some new strategies involve a re-taylorization of work or a transferral of autonomy from workers to algorithms.
Three different forms of control can be differentiated: direct control, indirect control and algorithmic control.
Direct control is exercised by superiors and it is based on direct observations and performance. Still machines can be used for measurements and exercising control. A superior who determines the speed of an assembly line is exercising direct control even if it is mediated by objects.
Peters and Sauer (2005) see indirect control as a new form of domination because it uses the autonomy of workers for control. In Labour Process Theory, Friedman (1977) already described this new form of control as ‘responsible autonomy’. Management just defines certain goals and sets certain conditions (technical supplies, target agreements, strategical priorities, resources, etc.).
Workers then have to define tasks to realize the goals and carry them out. Instruments for indirect control are for example management ratios, management by objectives and the use of technical requirements. A central aspect of indirect control is the use of market imperatives. Market imperatives can either be artificially implemented in the working environment (intra-enterprise competition) or actual capitalistic market imperatives can be passed through to workers. Then workers are supposed to identify with the economic success of their employer, develop an entrepreneur perspective and even set goals themselves that formerly the management had to define. The workers have to solve the transformation problem themselves: ‘Workers are thereby made responsible for the translation of their own labour power into labour output’ (Ferschli 2017: 172). The decisions of management vanish behind objective forces. But as Ferschli (2017) points out, the antagonism between labour and capital does not vanish and capital remains in control. Surplus value is still exploited by the capital fraction and conditions (like profit rates that have to be achieved) and goals are set by the capital side.
Algorithmic control is compatible with direct and indirect control. Big data, new sensors, integrated systems and machine learning are supposed to allow a real-time control of labour processes and constant feedback loops. In the case of direct control, algorithms are used to inform management decisions or automatically enforce goals. But algorithms can be brought to a more interesting and effective use in the case of indirect control. They can have a similar function to the incorporation of market imperatives. Management decisions like what is the algorithm supposed to optimize, which algorithms are used and in which areas are they used are opaque. They are hidden behind algorithms as they are often hidden behind market imperatives. Furthermore, algorithms can be used to observe markets and automatically implement market demands in the workplace.
The loss of autonomy described up to now is at least one aspect of alienation. Using the concept of alienation, Marx is capable of showing relations between personal autonomy and political autonomy, while modern debates often focus on personal autonomy. In this context, I will also focus mainly on the loss of personal autonomy or heteronomy at the workplace. But connections to political autonomy will be drawn. The use of market imperatives and algorithmic imperatives in indirect control is one of these connections while the tendency of re-taylorization and algorithm mediated direct control in digitalized work environments deals mainly with personal autonomy.
Working contracts
It is a notorious problem to distinguish between a contract for services and a contract of services (contract of employment). Only in the second case, an employer-employee relation exists and, therefore, the according rights of employees. We can already find this distinction in Roman law between locatio conductio operarum (contract of employment) and locatio conductio operis (contract for services). But since the ancient times, technology allowed us to blur the line between these two to a much higher degree. German settled case-law has defined for quite a long time attributes that define employment relations. In April 2017, legislation codified this understanding in §611a BGB (Civil Code):
The employment contract obliges the employee to performance of heteronomous work bound by instructions in personal dependence in the service of another person. The right to give instructions may relate to the content, execution, time and place of the activity. Those who are not essentially free to organize their work and determine their working hours are bound by instructions. The degree of personal dependence also depends on the nature of the activity. In order to determine whether an employment contract exists, an overall consideration of all circumstances is required. If the actual execution of the contractual relationship indicates that this is an employment relationship, the term in the contract is irrelevant.
The employer is obliged to pay the agreed remuneration.
Often companies try to argue that people are self-employed, even if they are given instructions about content, time and place of the activity. People can individually sue for employee status but often they will not do that because they might loose their job, it is costly or it takes too much time. The rise of a platform economy where services are only mediated between provider and buyer is quite problematic because workers who are in need of protection do not have many rights. Therefore, they often have a low income (sometimes far below rates they would get as an employee) and suffer from precarious living conditions. These self-employed workers often have to fulfil certain standards to use the platform. These standards could be understood as instructions as therefore ground an employment relation. But in many cases, this might be too far stretched. Another way must be found to give dependent self-employed workers the same rights and protections employees have.
For employers, a basic problem of employment contracts is how to transform the abstract labour potential of the employee into concrete measurable labour (Edwards 1979; Marrs 2010). This problem does not – or at least not to the same degree – occur in contracts for services. From the perspective of Labour Process Theory, this would justify a separate article. In this article, I will mainly deal with employer employee relations. The scientific management of Taylor is one solution to the transformation problem. As we will see in the next section, digitalization and retaylorization is a modern technological answer to the problem that involves workload compression and a loss of autonomy for the workers. Braverman (1974) analysed Taylorism in Fordism. Labour Process Theory needs to actualize this critique in the context of digitalization.
Digitalization and heteronomy
Often digitalization is propagated as extending autonomy. The scientific board of advisers of the Platform Industry 4.0 (Wissenschaftlicher Beirat Industrie 4.0, 2014) claims that many possibilities for self-organization and autonomy will emerge and new action opportunities for the workers will occur. But at least three effects are connected to digitalization that tend to foster heteronomy: formalization, data centred decision making and flexible structures.
Friedman (1990) saw two different managerial strategies to transform labour power into labour: Direct control strategy and responsible autonomy strategy. He argued that both strategies have problems. Direct Control involves unskilled workers who cannot react to changes in production and be easily transferred to different workplaces. Responsible Autonomy involves skilled workers who have bargaining power, influence at the workplace and cannot be easily replaced or fired. But as we will see, digitalization allows the flexible use of unskilled workers and the control of skilled workers. Therefore, digitalization (as a managerial strategy) is a new and more effective way to control the labour process without some of the old problems.
Digitalization as formalization
Christiane Funken and Ingo Schulz-Schaeffer (2008) were describing digitalization as a generic term encompassing all forms of operational use of information and communication technologies – their use as instruments of any kind of pre-structuring of operational processes as well as their use as a shapeable medium of communication. Kleemann and Matuschek (2008) highlighted that work is usually split up into formalizable, logical, mathematical dimensions which is recomposed in the production process. The use of machines and technologies for keeping track of formalized logics and quantified outcomes of work for management control over workers, is not new. Taylor and the Gilbreths used early machines and tools to monitor workers’ productivity, including michrochronometers, stopwatches and stethescopes. What is new is the use of cheap new sensors, robotics, big data analytics, artificial intelligence, cloud computing and communication technologies, which allow the widespread application of digitalization for control of the product life cycle and working conditions (Boes et al. 2017; Moore 2018). Digitalization of work, therefore, must be understood in the context of the formalization of work processes, including their fragmentation into single processes, their measurement and their conversion in mathematical numbers and formula. This includes control, evaluation, structuring and optimization.
Instructions given by the supervisor (based on working contracts) normally leave room for decisions by the workers how to fulfil a task. Digitalization formalizes working processes to a high degree and detail. Formalization fixates and standardizes rules for behaviour, processes and procedures (Funken & Schulz-Schaeffer 2008: 13ff). This has the consequence that unsystematic activities, diverse practices and informal processes and communications that have been done in many different ways and left room for autonomy are now registered and assessed based on efficiency criterions. Then working processes are restructured and optimized. Funken and Schulz-Schaeffer (2008: 12) also argue that there is a limit for formalization because workers have to be able to react flexibly in order to fulfil their tasks. But the use of big data and artificial intelligence allows digitalization and software to react flexible and give flexible instructions to workers by using assistance systems. Therefore, formalization is possible to a higher degree while at the same time a higher degree of flexibility is achieved.
The formalized standards are also implemented in software and this software demands specific inputs. Because the software allows only specific inputs, it also defines possible work steps. In medical and health care, for example, interactions between carer and patient are then standardized without the possibility of humane deviations. At the same time, the software always measures and controls whether the work steps are done within the defined parameters. An extensive control of workers performance is the consequence (Mengay & Pricelius 2016). If seen in the context of working contracts, digitalization extends heteronomy because the instructions for the execution of work steps that are given as a consequence of the formalization process or just by the digital visualization of processes in process descriptions are binding for dependent employees. Deviations from digitally visualized working procedures are not possible or are violations of working duties. Deviation to the digital standardized work requires justification by the employee and can result in dismissals.
Sometimes, this heteronomy is to a certain degree concealed because it seems to be just a matter of software inputs or the logic of work organization. But implementing digitalization, standards, software and new forms of work organization is still a decision of management to keep profits high and to control the work process. Windelband (2014: 157) sees this as new form of digital based Taylorism in which the worker is alienated from the working process because of the virtualization of these processes. Even the smallest escape methods to keep moments of relaxation and autonomy during work can be identified and rationalized away. Moore and Robinson (2016) show how even emotions, health status and stress levels of employees are measured to optimize work output. While even big tech companies tried to control character, emotions, stress, health and motivation by non-digital technologies, new developments allow digitalization also to extend the access to the inner autonomy of workers. The degree of heteronomy will be higher in unskilled labour than in skilled labour, but with digitalization even skilled labour will be heteronomous to a rising degree.
Data centred decision making
Digitalization allows to extend heteronomy by collecting data about working processes, analyse it and give instructions based on the analysis. Of course, the collection and analysis of production data is nothing new. But digitalization allows to collect an amount of data (big data) that can only be analysed with modern computers and algorithms. Data collection and analysis is also extended to dependent self-employees, offices, hospitals and other service providers, including state run services. In general detailed measurement, comparison and visualization of performance or productivity allows to put a lot of pressure on employees to work more efficient and changes the character of work. As a result, digitalized work becomes more and more an appropriation of the work concept in itself and facilitates a rearticulation of work as performance. Together with the formalization of work steps, the analysis of the acquired data can be used to control and optimize every small activity of the workers and to improve the company performance. The data based comparison between workers and work organization can be made on a global scale. Therefore, performance control and work intensification is a part of the digitalization of dependent work and the result of a violation of data protection rights of the workers.
Often data collection is made in secret so that workers first have to be able to identify which data are actually collected. But there is not only a transparency problem concerning the collected data. It is also normally not transparent on which basis the collected data are evaluated. Therefore, it is harder to criticize the management decisions if the workers do not have access to the data and the evaluation criteria.
A further problem occurs if data based decision making is automated (Mader 2018). Artificial intelligence can evaluate the data and optimize the work flow. In this case, it might not even be transparent to the management on which basis the automated decision were made. But the management still defines the parameters that the algorithm is supposed to optimize.
Flexible structures
The flexible structures that are often part of a digitalization will be used in a way that is advantageous for the people with more power. In times of weak labour unions, this will be most likely the employer. Instead of using flexibility for the autonomy of the workers, workers will have to follow very flexible instructions concerning time, amount, place and kind of work.
This data analysis fits well with concepts of lean management and agile production. Work packages can be flexibly distributed to people who will complete them most efficiently (according to the data). In methods like scrum or Kanban working teams often have to decide themselves how to handle the tasks and who will do what. Based on data analysis and prediction methods work package distribution can be based on algorithms. Autonomy of individuals and teams are lost. In extreme cases like crowdworking or dependent self-employees, labour can be used based on pull-principles like in the cases of resources.
Reference architectural model industrie 4.0 (RAMI 4.0)
RAMI 4.0 is a reference architectural model Industrie 4.0 designed by the Plattform Industrie 4.0 (2016) intented as a standard for a common communication structure about how to realize digitalization in especially industrial environments and later standardized as a norm and technical specification for Industry 4.0 in the DIN SPEC 91345:2016-04 (2016). It is an example for a technology and profit centred model that is not interested in working conditions or workers co-determination.
Therefore, the life cycle and value stream are integrated in RAMI 4.0 and connected to production hierarchies and digitalization layers without considering working conditions.
(RAMI 4.0 Source: Plattform Industrie 4.0 (2016)).
A digitalization project that follows the RAMI 4.0 standard will be bound by the RAMI 4.0 structure. Workers co-determination, higher wages or better working conditions could be seen as a deviance, as a problem for the life cycle or the value stream. It is an example of how formalization sets standards and norms. Only selective data are measured, evaluated and used for the digitalization project. The political questionable intentions of life cycle and value stream as well as forms of hierarchy are integrated. Humans also become assets that are digitally represented (Moore & Robinson 2016) and integrated in a human-machine system that is regulated by algorithms to optimize profits.
RAMI 4.0 includes different layers that represent the digital integrations of assets. The material asset (e.g. component, screwbox, a plan, etc.) is translated or virtualized into a non-material representation on the integration level. This includes an interface for data exchange. The communication layer consists in data exchange on the basis of an industry 4.0 communication protocol. In the information layer, real-time data allow to aggregate new information. In the functional layer, services and functions are available and can be applied. In the business layer, services and functions that are necessary for business models are provided.
To realize a digitalization project, several steps have to be followed. First, the assets for digitalization have to be selected and resources for the project have to be specified. Then, the current state has to be identified (layer, hierarchy level, life cycle value stream) and then the target state and goals for the digitalization project have to be defined. The change to predictive maintenance necessitates for example that data are not analysed by humans but by algorithms. Real-world objects as material assets become digitally doubled and identifiable, for example, with an RFID chip that identifies the object and allows assigned data collection. Databases collect and store the relevant information about the asset and its interrelation with the world on the information layer. On a functional layer, maintenance can automatically be initiated based on predictive patterns won through big data analysis and statistical experience with other assets of the same type.
Assets are contained in an administration shell. It stores all data and information about the asset. The administration shell functions as an interface for industry 4.0 communication, provides a virtual representation of the asset, contains a resource-manager, a manifest which includes meta-data and necessary industry 4.0 information and functional aspects like software and configurations.
In the RAMI 4.0 model, it is obvious that the management defines objectives, assigns resources, and defines goals and positive outcomes. Furthermore, the data that are contained in the administration shell is quite selective. It is declared that all information is included. But only data are included that is necessary for technical optimization processes or selected by the management or technical experts. Data about the quality of the workplace or economic democracy are therefore generally missing. To change that it is not sufficient to change something at a single administration shell or factory but the standards of industry 4.0 would have to be changed. The technical description of RAMI 4.0 veils domination relations moulded into formalizations and the social impact of restructuring by digitalization.
Conclusion
While the automation of production and the proliferation of integrated production systems (lean or Toyota) since the 1980s has significantly reduced, the autonomy of those employed in production, the new wave of digitalization of work is increasingly targeting the remaining field of autonomy (mostly of employees). Digitalization is affecting also good jobs in front and back office, administration and development. Digitalization enables a restructuring of the work content with improved possibilities of surveillance and control. Targets of most digitalization projects are workload compression combined with performance improvement complemented by technological heteronomy.
Digitalization of work produces new losers and collective experiences of heteronomy that could be the basis for new forms of resistance and collective struggle. The loss of autonomy can be seen as a shared experience of mis-recognition (Honneth 1995) that leads to struggles for recognition. A new relation between workers who are fully controlled by automation and employees who are fully controlled by digitalization could emerge. The common experience of technological heteronomy could be the motivational base for shared struggles against the heteronomy of work.