Keywords

1 The Rise of Digital Work

The general term “digital work” refers to work activities that are largely based on a digital infrastructure, are integrated into it, or require the use of digital tools in core areas of value creation. Key technologies are sometimes identified as drivers of digital work. The World Economic Forum (WEF 2018, pp. 5–6) identifies four technological drivers:

  • Robotics

  • Mobile/Social Media

  • Internet of Things/Connected Devices

  • Cognitive Technologies (Artificial Intelligence and Big Data Analytics)

Digital work is found in both the production and service sectors where the mentioned technologies offer different significance and distribution. Broadly speaking, it can be said that the digitalization of work evolves essentially around the integration of the internet with its integrated technologies in value creation processes. In addition to changes in existing tasks, new job profiles and occupations are emerging as well. As in earlier transformation processes (e.g. Womack et al. 1991), work is not distinctively determined by these technologies. Rather, different scenarios are forming that concern two central aspects (Gartzen-Wiegand et al. 2021): The automation of work through the use of technologies and the decision regarding human leadership versus mechanical guidance in digitized work processes.

The automation potential of digital technologies has been studied scientifically for some time (e.g. Autor et al. 2003), but particularly became the focus of public attention in politics and the media with the methodologically innovative publication by Frey and Osborne (2013). The authors studied the existing potential of digital technologies to substitute human labour and estimated it at 47 percent of all labour activities in the US. The focus is not on actual substitution, but rather on the available substitution potential at a certain point in time or within a certain period of time. This forecast led to intensive debates and a sustained discussion, especially among labour market researchers regarding the “potential for substitution” of digital technologies.

For example, in Germany, the Institute for Labour Market and Employment Research of the state-run Federal Employment Agency publishes regular studies and follow-ups on the topic (e.g. Dengler and Matthes 2021). This shows an ongoing dynamic, as new technologies continue to become marketable and thus have to be added to the potential. According to this latest study, the share of employees working in an occupation with a high substitutability potential is now around 34%, compared to 25% in 2016. The potential is considered “high” if at least 70% of the activities in the occupation are potentially substitutable. Dengler & Matthes conclude that even “more complex activities ... can be increasingly automated” (ibid, p. 1).

It can therefore be said that a considerable and increasing proportion of existing activities within occupations can be automated disregarding those occupations and activities where digitization is fundamentally impossible or undesirable. Nevertheless, occupations and activities remain that are digitized, and the question arises of what role man and machine in the respective work hold. Machine guidance is given when the collection of information for the execution of a task is determined by a machine, the solution path is predetermined, and decisions are made automatically. An example of this would be the repair of machines, where the diagnostic data must be partially entered manually by the human and the actual repair must then be carried out by a person according to specifications, while everything else lies within the machine. This can lead to de-skilling of job holders if comparatively demanding tasks that were previously the responsibility of humans are now being automated.

However, it is just as possible that this creates access to work for low-skilled workers who would otherwise have found it difficult to enter the labour market. On the other hand, human guidance is given when the selection of the relevant information sources as well as the final decision on which solution path to take lies within the human being and the machine functions as support. An example of this is a digital dashboard for the presentation of (real-time) information on an issue, the information selection of which is freely configurable by the user, within certain limits. Gartzen-Wiegand et al. (2021) evaluated existing publications on human-machine interaction scenarios and came to the conclusion that human guidance can be found at all skill levels, from assistants to experts, but is more likely to be expected with increasing skill levels.

Exclusively those activities, in which humans take the leading role, are considered digital work in this article. Equipped with digital technologies, but in the leading role over the machine, these activities are characterized by a fourfold “unleashing” compared to pre-digitalization:

  • Information: Rapid access to a wide range of information, its storage, filtering and visualization, statistical processing and simulation options for trends multiply the amount of information available at any given moment and can also significantly improve the quality of information, for example through sensor technology and automatic processing of information. After a work environment with as little of already standardised information as possible, we are now entering a working world characterised by information overload.

  • Autonomy: Comprehensive information in connection with sufficient qualification enable us to make decisions on the spot and thus avoid lengthy and error-prone processing procedures. This requires suitable scope for action and decision-making, specifically the allocation of corresponding competences to the persons carrying out the tasks. Through digital networking, their decisions can also have significantly more far-reaching effects on other, even more distant areas in their own organisation or other organisations.

  • Performance diversity: such work activities allow the use of a wide range of individual skills. Individuals differ not only in cognitive abilities, but also in personality traits such as creativity, openness, or conscientiousness. In work activities with multiple sources of information and high levels of autonomy, these differences mean that performance can no longer be considered normally distributed across different job holders. This leads to a curve more reminiscent of the Pareto distribution instead of the symmetrical distribution of performance around the mean. Here, most jobholders are similar in their performance, while a few individuals exhibit a far greater performance (Mühlbradt 2020).

  • Growth: Digital value creation processes, especially in the field of immaterial services, operate at least partially with available technical infrastructures free of charge, open-source solutions and inexpensive devices. Large investments are no longer necessarily dependent for high level technology in order to enable innovation and a high degree of utilization. Building on this, these processes have an enormous growth potential. Taking the freight carrier business as an example, if the process of daily routeing of the trucks could be automated via artificial intelligence (AI) so that time for human planning decreases from hours to seconds, more human capacities would be available elsewhere. Therefore, there would be no reason why this capability should not be made available via internet for thousands of freight carriers.

The following discussion demonstrates that our capabilities to design unleashed digital work with the dynamics of change need to keep up with the time. This requires a new way of thinking that is closely linked to our understanding of complexity and how we deal with it.

2 Complex Sociotechnical Systems

Groundwork on systems theory has been published by Bertalanffy (e.g. Bertalanffy, 1940), a highly influential modern theory. Although some of the roots go much deeper (Lutterer, 2021), this consideration focuses on this work. Likewise, we will not go into the numerous perspectives and applications of systems theory but limit ourselves to the field of work analysis and design. This refers essentially to the disciplines of work science and work management, work and organisational psychology, industrial sociology, and the interdisciplinary field of human factors. Today, we usually speak of “work systems” in these mentioned disciplines.

In general a “system” consists of several interacting elements. Systems figure before a background called environment. Systems are partially open and interact with their environment through a boundary. Furthermore, they are self-regulating and based on feedback. This abstract concept of a system is suitable for any sections of reality that makes it a meta- or universal theory for application in different sciences and realms of reality. Figure 1 shows the two standard views of an abstract system.

Fig. 1.
figure 1

Two views of abstract systems

Full size image

The left illustration shows a frequently encountered representation (e.g. Schlick et al. 2018, p. 22). A system interacts via input and output with its environment, from which it is separated by a partially permeable boundary. Subsystems or elements within the system can be connected to each other via relations. The emphasis is on the structural side and is therefore often selected to represent objects “systematically”.

Quite different, but completely equivalent, is the representation of a system as a chain of elements acting on each other, so forming a process. The system in the right illustration consists of the elements A-D. Everything else belongs to the system environment, which one has to imagine as an amorphous background (e.g. Senge 1990, p. 97). The focus is clearly on the process, where the elements can be concrete activities, for example a number of activities as in making coffee. However, they can also be impact factors or states. This form of representation is particularly suitable for constructing models to explain phenomena.

A “systemic” approach focuses not on isolated elements, but instead on the system as a whole. Thus, the concept of “emergence” refers to a class of phenomena (e.g. system properties or actions) that are caused excessively by an interaction of elements that ultimately cannot be clearly determined, in accordance with Aristoteles expression “the whole is more than the sum of its parts”. Admittedly, the system can be broken down into parts, however, this just does not explain the emergent phenomenon in question.

Applying the concept of systems to human work in connection with technology, an evolution of the understanding and conceptualization of systems can be demonstrated. For the production and use of simple tools, it is sufficient to consider the object itself.

The use of machines as technical entity with moving parts and propulsion systems does not represent qualitative changes as long as the propulsion can be rendered through draft animals, wind energy and hydrodynamic power. The design of technology remains relatively unproblematic if these machines do not provide large amounts of energy and thus cannot deliver high velocities. This changed with steam power, the internal combustion engine, and the electric motor in combination with further technical inventions. Power, velocity, and risks far beyond the usual “human scale” made a broader perspective necessary. A leap of innovation in this regard was initiated by the Second World War. For example, the speed of an aircraft increased by 300 percent and the number of controls and instruments in the cockpit by 350 percent (Badke-Schaub et al. 2012, p. 5). Considering this background, Hollnagel (2021) sees the advent of the man-machine system (MMS) concept around 1945. Under the impression of increasing technical possibilities, the MMS became a concept and object of design. The concept is that both elements - man and machine - must be considered in their mutual relationship so that the totality of both can operate successfully.

A little later, under the influence of a field study in British coal mining, the concept of the “socio-technical system” (STS) emerged (Trist and Bamforth 1951). After a change in the technical method of coal extraction in a mining company changes in tasks as well as responsibilities were experienced as deskilling by the miners. The concept of STS extended the human component of the MMS from individuals to teams and demanded a conscious design of the social subsystem as well as the joint optimization of the technical and social subsystem. This insight triggered a continuing preoccupation with work organization in occupational psychology (Ulich 2011), industrial sociology (Hirsch-Kreinsen 2014), and labour science (Heeg 1991). In these days, the term STS was almost considered a conceptual common good, which has become an integral part in the concept of Industry 4.0 in Germany (Kagermann et al. 2013).

With increasing frequency and not just recently, work systems have been associated with the term “complexity” or the adjective “complex”. The following statements cover a period of about 40 years from the “Ironies of Automation” (Bainbridge 1983) to the WHO’s Patient Safety Action Plan (2021), without claiming to be - even approximately - complete. The definition of “complex systems” range from a self-evident term that does not even need to be explained (Grote and Kolbe 2015) to a particular discussion of the term complexity (Jenkins et al. 2009; Latos et al. 2017; Patriarca 2021). Different, sometimes interwoven, views of complexity can be found.

One view is the complexity as a property of a system (Latos et al. 2017), while process complexity (Bainbridge 1983; acatech 2016) is considered a linguistic variant of this. Jenkins et al. (2009) gave their work on the methodology of work analysis with the significant title: “Cognitive Work Analysis: Coping with Complexity” and considered the complexity of systems to be gradual. As essential properties of complex systems they looked at system dynamics, the number of components and relations, the predictability of future system states and the extent of risks through actions.

Another point of view focuses on the environmental complexity faced by work systems (e.g. Walker 2015; Meissner and Heike 2019) or “real world complexity” (Dekker et al. 2008).

The WHO uses a combined perspective in the Global Patient Safety Action Plan (WHO 2021). In it, the healthcare system is described as: “complex amalgam of actions and interactions, processes, team relationships, communications, human behaviour, technology, organizational culture, rules and policies, as well as the nature of the operating environment” (WHO 2021, p. 2).

Another perspective arises from Schaper (2015), he speaks of “highly complex services” that are being provided in the health care sector. This corresponds to complexity as task complexity. A comparable view is employed in Norman (1986) concerning the use of devices.

In summary, complexity is understood as a property of the system, the environment, the output, or all of these together. This status quo is to be criticized in two respects. First, there is a confusion regarding complexity and complicatedness: a large number of interacting parts is not complex if this does not play a role or poses a risk. On the other hand, Norman (1986) mentioned a vivid example in his introduction where novice sailors were already overwhelmed when performing using a single parameter (steering a sailing ship according to a compass). Secondly, the relationship between objective and subjective complexity is unclear. Is complexity objectively present, or is it experienced differently between individuals? Patriarca (2021), for example, concludes that complex systems are those that are never fully understandable (“knowable”) and states regarding the resilience engineering approach: “complexity is not considered a thing per se, rather it is a situation to be investigated” (ibid, p. 479).

Latos et al. (2017) also point out that human perception of complexity is also subjective. Norman (1986, p. 33) analyses the concept of task complexity and concludes: “the correct conceptual model can transform confusing, difficult tasks into simple straightforward ones”. This anecdotal sequence is complemented by a systematic literature review of long-term trends in research on socio-technical systems (Mühlbradt et al. 2022). The analysis was part of the session ‘Current approaches to the analysis of complex socio-technical systems’ at the Spring Congress 2022 of the German Society of Ergonomics. The aim was to systematically display research trends considering methods and designs of socio-technical systems based on the date of publication using SCOPUS literature database. Articles were included that contained a combination of the words socio-technical and system/approach/design/method/analysis in German or English in the title, abstract or keywords. 5,664 publications from 1967 until 2022 were found. In addition to thematic clusters, three hypothetical trends alongside the increase in complexity of sociotechnical systems emerged (Fig. 2).

Fig. 2.
figure 2

Hypothetical trends based on the literature review (Mühlbradt et al. 2022)

Full size image

The authors found a development in publications that increasingly refer to the term “complexity” and either postulate an increase in complexity or argue for a stronger awareness of already existing complexity in procedures and methods. Three focus areas can be identified in this development:

  • Designing frameworks and sophisticated methodologies for designing IT and/or work systems under the heading “systems engineering” (e.g. Leveson 2012).

  • The resilient organisation as a mission of socio-technical design with a focus on the health care sector (e.g. Hollnagel 2018).

  • Transfer of socio-technical approaches to regional and substantial societal macro-systems (transport, urban planning, etc.) far beyond work systems (e.g. Gebhardt and König 2019).

Since Bainbridge already discussed the effects of complexity in the context of work systems in 1983, publications on this can be found continuously and one could initially assume that this has always been an aspect of work systems. In this respect, it is possible that individual differences can be found. In this case, the individual inability to understand a system (an environment, a task) should be remedied by human resource management measures such as personnel selection or personnel development.

However, if complexity is present for most or all people in the system, this strategy fails. It is argued here that the factors

  • digitization of work

  • unleashing of work

  • increasing demands on performance and efficiency of processes

  • increasing regulation

come together and lead to more and more work systems, tasks, processes, or environments that can be perceived as complex. The phenomenon of complexity is thus changing from an exception that affects a few individuals towards the fact that it affects many or all people concerned (employees, manager, customer, partner). This on the other hand leads to a crisis of work design, as will be argued in the next section.

3 A Crisis of Work System Design

Work systems change with their requirements and the available technical possibilities. With them, paradigms and methods of work system design must change, as already stated above. The question arises whether methods for designing complex socio-technical systems are available and how good these methods are. For a closer inspection concerning this question, the health care system is a good place to start. This is because, according to the WHO, it is a “complex amalgam” (WHO, 2021, p. 2), or, in the words of Braithwaite et al. (2017, p. vii): “many people believe that healthcare is the example par excellence of a complex adaptive system”. Within healthcare the focus of work analysis and design is often on the criteria of patient safety, the prospective avoidance of possible errors and the analysis of past errors. This can be considered a special case of work analysis and design.

Braithwaite et al. (2015) claimed that healthcare is more complex than linear methods of find-and-fix error analysis would allow and judge: “We have to acknowledge the intricacies and complexity of healthcare to overcome this limitation” (ibid, p. 419).

Schrappe (2018) reflected on the improvements achieved in patient safety in Germany with reference to the guiding publication “To Err Is Human” (Kohn et al. 2000). He aimed to make reliable quantitative statements and concluded that a range of 400,000–800,000 adverse events happen in hospitalized patients including an avoidable mortality rate of 20,000 patients annually. Considering these numbers, he called for a revision of “cherished views” (op. cit., p. 353). The concepts of complexity and complex systems play a particularly decisive role (op. cit., p. 7).

Wears and Sutcliffe (2019) also find that too little progress has occurred within the past 20 years. They blame the inefficiency of safety efforts and demand a radical reform.

A bibliometric analysis on the occasion of the twentieth anniversary of “To Err Is Human” for the time period of 2000–2019 was conducted by Pierre et al. (2022) that included approximately 20,000 publications that referred to “To Err Is Human”. The authors concluded that the main emphasis on traditional approaches regarding risk and error management is too excessive and a lack of systemic perspectives exist. Although these systemic perspectives were included in the original publication, they were hardly to be found in subsequent works. The authors are in line with Schrappe, who also lamented the lack of “concepts such as systems or complexity theory” (Schrappe 2018, p. 66).

Furthermore, Sujan et al. (2022) found that actual improvements remain low of expectations and ascribed this to the exclusive use of simplistic and reductionist approaches.

Overall, there is increasing dissatisfaction with the status quo and also with the paradigms and methods of work and process design that are used. An explanation for this dissatisfaction is consistently related to theoretical-conceptual as well as methodological deficiencies. To put it bluntly, the deficits can be interpreted as dead ends in coping with complexity. Particularly, there are four such dead ends:

  • Complexity as a feature of something. According to this notion, something is objectively (measurably) complex. A confusion regarding the concept of “complexity” often occurs, which can be understood as a number of elements and their relations that can be expressed by a number. However, all elements and relations of complicated systems are known.

  • Decomposition of the system in the analysis until there is no more complexity. This analysis precedes the synthesis, in which the phenomenon of complexity no longer occurs.

  • The “Root Cause Analysis” (Badke-Schaub et al. 2012, p. 16) the search for causes of error. The ultimately found cause is isolable by definition from the system and the relationship between the root cause and the revealed error is clear and explicit.

  • Systems engineering tries to design and develop methodologies that can handle complexity. However, this also does not lead to a disappearance of complexity, but rather shifts it into the methodology and the interpretation of the results so achieved.

A pause before further argumentation is appropriate at this point. There are two arguments against a crisis in work design; the first argument relates to the problem of complexity.

If important goals such as patient safety are not satisfactorily achieved because of difficulties in analysis and design of complex systems and the work within these complex systems, then why not design simple, linear systems that spare complexity? The answer to this question has already been given by Charles Perrow (1984, p. 89): “we have complex systems because we don’t know how to produce the output through linear systems”. Linear processes can run stable, efficiently and can be permanently maintained if the environment in which they operate is either simple or has been simplified beforehand by human intervention. The well-known example of a simplified environment is Henry Ford’s famous saying that you could have your car any colour you want as long as that colour is black. Our goals and our demands on work systems primarily lead to complexity. There is no other way to generate “complex performance” (see above) when pursuing highly set goals and diverse requirements, than via complex socio-technical systems.

The second argument questions the need for an explicit design of complex work systems. According to this viewpoint, largely autonomous, highly qualified job holders would be able to make the necessary decisions on their own or in a team. Sufficient degrees of freedom at the workplace and available resources would be merely necessary. In fact, there are numerous references in recent work design methodologies regarding employee participation in work design, ranging from lean management (e.g. Bertagnolli 2018) to the design of software tools and user interfaces (e.g. Steimle and Wallach 2022) to setting performance goals for work groups (e.g. Debitz et al. 2012). However, a complete abandonment of work design would ultimately lead to a lot of autonomous subjects that would have to coordinate themselves independently. The costs of this would be high in a digitalized world with multiple networks (Barabasi 2002). To support the individuals in this respect, an organisational framework is helpful. Moreover, the above-mentioned examples of participation-oriented methods do not intend to provide an overall picture of the end of work design. Rather, a permanent change is taking place that involves tasks, forms, and instruments of work design.

Finally, it should be mentioned that the statements regarding the healthcare sector can be generalized. The healthcare sector is a “laboratory of complexity” due to special requirements and framework conditions in which numerous circumstances come together. However, it is by no means an incomparable special case.

In order to get to the bottom of the necessary changes in philosophy and methodology of work analysis and design, it is necessary to temporarily leave the field of work behind and dive into general psychology, or more precisely, into problem-solving research. Here, complexity does not present itself as a property of systems, but as a human experience.

4 Complexity as a Borderline Experience

Borderline experience can be defined as an event that is subjectively experienced as chaotic and exceeds one’s own possibilities to create or find transparency and order. Under such circumstances, humans find themself at the limit of their competence to predict and control reality. In Germany, the beginning of a public discussion of systemic effects can be associated with the publications of Frederic Vester (e.g. 1974). Such research often relates to the term “systemic thinking” (Lutterer 2021). In distinction to this, approaches to work design tend to be considered “systemic”.

The central idea of systemic thinking is that simple linear causal relationships (A causes B) are often illusory and that it is much more important to assume a multitude of interconnected and mutually influencing factors. Some, or even many, relevant factors and the nature of their connection to other factors (promoting or inhibiting, rapid or delayed effect, etc.) may be unknown at a given point in time. If such hidden side and remote effects cannot be taken into account, implications for operational risks arise.

The field of cognitive psychology investigates thought and cognitive processes in humans. Part of it is since the beginning of the 20th century the study of problem-solving behaviour in humans and animals. The general definition of a problem can be that it is a task that requires the transition of an initial state into a target state whereby the appropriate approach is unknown at the beginning. The distinction between simple and complex problems (Öllinger 2017, p. 590ff) describes a significant differentiation. Simple problems are static, have clear initial and target states and the available options for action (“operators”) are basically known. Complex problems, on the other hand, are dynamic (self-changing) and have a large number of interconnected variables (influencing factors) whose effects and interactions are at least partly unknown. The scientific study of such complex problems requires an affordable and effective technical platform, which became available with the personal computer. Therefore, people’s engagement with constructed situations was studied in the psychology laboratory soon after (Dörner et al. 1983; Dörner 1989; Funke 2012). In this process, people are exposed to a computer-simulated reality in which they have to act in a goal-oriented manner, although their understanding of this reality is limited and partly flawed. Their task is to achieve certain goals in the simulated reality over a period of several hours. For this purpose, they can intervene in successive game cycles by means of known options, in order to then experience the reaction of the simulation to this intervention and, if necessary, change their strategy as a result. However, the number and linkage of the factors in the simulation is not known at the beginning and must, ideally, be developed independently throughout the course of the simulation cycles. In other words, the test person must solve the problem of absence of knowledge using correct, i. e. goal-oriented, intervention behaviour.

The simulations are semantically associated and assign a role to each participant. Figure 3 shows one aspect of such a simulation. In this case, the participant is an aid worker and has to increase the survival rate and prosperity of local inhabitants in a savannah region. One way to intervene is to drill a well. This increases the amount of available drinking water, which in turn boosts the cattle population and thus contributes to prosperity. The idea of drilling more wells leads to success - a self-reinforcing cycle. However, the groundwater level is another unknown variable hidden in the simulation. It is lowered by the water withdrawal - which has no consequences as long as the level does not fall below a critical threshold. But following, the amount of available drinking water decreases abruptly. Such progressions are typical for exponential functions (see the small box at the bottom).

Fig. 3.
figure 3

Knowns and unknowns in a laboratory problem-solving task

Full size image

In simulated scenarios, participants very often fail to derive and apply an adequate mental model, so that their interventions fail or produce overwhelmingly negative side and remote effects. Consequently, the attempt to achieve the set goals cause frustration and probably an even increasingly worse experience.

Therefore, concerning these simulations one does not like to speak of complexity, the word that comes immediately into mind is “chaos”. This ancient Greek word originally denotes a wide empty space or a gap and stands today for complete disorder or confusion. To fight this confusion, no ordering and action-guiding model can be established. The result is either the experience of one’s own helplessness or the stubborn insistence on wrong models and strategies. Considering the second case, the eventual collapse of a system will often be explained by factors beyond one’s own responsibility.

High-level problem solvers differ less from low-level problem solvers in terms of their intelligence - although, contrary to earlier assumptions, intelligence is also relevant (Leutner 2002; Weise et al. 2020). Dörner (1989), however, finds primarily certain behavioural aspects in high-level problem solvers. They are characterized by advanced levels of information search and integration (i.e. model building), elaboration and balancing of goals, making predictions about developments and a pronounced self-management to deal with stress and frustration.

According to Tatlock’s findings (Tetlock and Gardner 2016), high-level problem solvers are:

  • open, curious

  • self-critical, cautious and humble

  • they take nothing for granted, reality is complex for them

  • they are good with numbers

  • they think in terms of probabilities and are always ready for an update

  • they want to improve, and they have “grit”

The best predictor of performance is the degree of willingness to give up personal beliefs and the will to improve. This is about three times more important than the second-best predictor, intelligence. The best laypersons (named “superforecasters” by Tetlock) were superior to experts.

Laboratory and field experiments deliver consistent results to where the interaction of intelligence and behaviour determines the outcome of the experiment. What seems to be decisive in both cases is to be aware of complexity, to presume a lack of knowledge, the need for further hypotheses (models) about reality and the persistent assumption that one could also be wrong. Complexity does not just disappear. However, progress can be made.

5 An Integrative Approach

The fact that complex socio-technical systems are not simply complicated systems cannot be circumvented by methodological techniques, as discussed above. Approaches to eliminate complexity through breakdown and looking at isolated components do not lead anywhere, because they ignore Aristotle’s insight that the whole is more than the sum of its parts. However, methods also lead to dead ends that attempt to do justice to complexity through an extensive methodology.

These methods might work in theory, but when practically applied they overwhelm human beings, since the methods and results themselves turn complex. Noticeably the following phenomena take place:

  • The methodology is not fully understood or at least not within normal boundaries of time and effort

  • The methodology does not produce results in a stringent and objective way but needs creative interpretation of results and imagination

Successful engagement with complex socio-technical systems must consistently integrate the findings of problem-solving research in order to adequately address complexity.

Both traditions must be brought together conceptually and methodologically: “systems” as objective conditions and “systemic thinking” as a subjective level. Particular methods and instruments are used that have such a double system-theoretical orientation within this framework as shown in Fig. 4.

Fig. 4.
figure 4

Integrative system-theoretical approach

Full size image

If socio-technical systems in work environments are expected to provide “complex services” in an unfiltered “complex environment”, then complexity will be experienced by people who act within the system. This relates equally to people working in the system and to people who cooperate with, design, or analyze such systems - there is no fundamental difference between them.

One is almost inclined to speak of virtues regarding the skills identified in psychological experiments. Therefore, the successful and high-level problem solver should also have appropriate competences for people who deal with work systems. And this is always the case when goals and requirements entail complexity. Dörner’s laboratory experiments and Tatlock’s forecasting studies show that lay people who have a certain level of cognitive capacity and specific virtues such as openness to criticism, a willingness to learn, grit and modesty can achieve very good results and outperform professional experts. There is no reason to assume that this should be different in work systems under the condition of complexity. Any encounter with a living work system in unfiltered (unleashed) circumstances can potentially be characterized by complexity. The study of Bos et al. (2022) is a good example of how fast this can be the case. They performed an investigation in a hospital to find out how a programme for follow-ups of discharged patients was actually implemented by the different departments involved and concluded that the programme was not implemented as planned in any of the cases. The practitioners’ process descriptions were more extensive than the planning model and different departments also differed in their practice. Managing complexity successfully starts with acknowledging it.

This integration is not synonymous with the abandonment of all efforts to understand and improve work and organization. However, the claim of complete analysis and design of complex socio-technical systems is very much replaced by the goals of modelling and developing these systems. Dealing with the borderline experience of complexity is not primarily done by applying existing expert knowledge, but rather by a learning process. In other words: The confession of one’s own unknowing, the will to understand more and the continued effort to do so, coupled with the ability to be self-critical, are crucial for progress. Within this process, just an improved preliminary result can be achieved, but not a final point.

The integrative approach also does not negate the appropriateness of simple approaches to simple linear systems. Methodologically sophisticated analysis techniques and clearly structured design methods can be suitable means in these cases. The price of a simple system is, if applicable, an artificially simplified environment. The question arises of how high the costs of implementing and maintaining this simplification in relation to the benefits are. A good example of rethinking simplification is the introduction of gain sharing for industrial work systems (Thomas and Olson 1988). Gain sharing assumes that work teams are able to find their own way without prior planning to make their work more efficient so that they achieve certain production goals more easily. If significant and stable improvements are achieved, the teams can then sell these to the company. The company then officially implements them and in return sets a higher performance threshold for premium payments. Turning away from simplification too late is reflected in practices that still work on paper but are undermined in reality (e.g. Siegel 2015). The question arises whether one considers a system to be a simple system only because one has formally classified it that way. Pfeiffer and Suphan (2015) argued that “simple routine work” in industry actually demonstrates a substantial range concerning the level of work contents and demands.

Finally, one should not be deceived by the successes of linear approaches if these successes are based on the sole consideration of selected parameters in a narrow focus, while ignoring potentially negative side and long-distance effects of the implemented measures (Woodnutt 2018).

One might also be tempted to object that the design approach of systems engineering criticized above has impressive successes to show. However, these successes are based on technical structures such as the most recent example of high technology, the James Webb Telescope. It contains many physical components and relations that are certainly very complicated, but not complex. This design task can be mastered with a suitable system, sufficient time, good training, motivation and team strength.

Nevertheless, these systems are not capable of dealing with unknown variables and their interactions and the systemic phenomenon of emergence. The best current approach in the technical direction is that of artificial intelligence research. Their currently dominant solution, the artificial neural networks, achieve their goals through pattern matching with numerous cycles of learning. Delivering a realistic view, Marcus and Davies (2019) ascribe the lack of representation of the world, the absence of causal models, and a “blank slate” view of understanding and learning to these approaches. It is the people who adapt their work to different and dynamic conditions. Therefore, socio-technical systems are still superior and different laws apply to their design.

The integrative approach can be concisely described by a series of axioms. These axioms apply to working persons in “working systems” as well as to their designers:

  1. 1.

    There are no systems in the world - except through us

    “Systems” are arbitrary or deliberate constructions of our mind that we use to better understand the worldFootnote 1.

  2. 2.

    Nothing is complex - except for us

    Complexity is what we experience when our abilities are overstrained and we therefore cannot fully describe, explain, predict or change something (cf. Winograd and Flores 1995). The statement that something is complex in itself is meaningless. Complexity can only be experienced in the context of a value- and goal-oriented human activity.

  3. 3.

    The ability to deal with complexity varies

    People, teams and whole organisations differ in their ability to deal with complexity. Personality traits as well as thought patterns and behaviour are important for this, but so are team and organisational cultures.

  4. 4.

    Modelling instead of analysing

    Complex systems can be modelled. Complete system analyses of complex systems that resolve complexity, on the other hand, are impossible because:

    1. a.

      Elements and states can never be reliably and completely specified

    2. b.

      Sufficient resources are never available

    3. c.

      If a) and b) does not apply, the results of the analysis would themselves be complex

  5. 5.

    Develop instead of design

    Learning with the assistance of appropriate models can lead to improvements in dealing with complexity. Learning progress enables the generation and the use of more sophisticated models in a sense of “absorptive capacity” (Cohen and Levinthal 1990; Braithwaite et al. 2021).

  6. 6.

    “Perrow clause”

    The only reason to design systems that confront us with complexity is that we do not know how we could achieve the same performance with simpler solutions (see Perrow, 1984, p.89).

The concept of the “resilient system” captures a prominent position in the integrative approach. The concept explicitly refers to complex socio-technical systems that are understood as not fully analyzable and predictable (Hollnagel 2016, 2018; WHO 2021). Resilient systems are those that are able to maintain their purposeful functioning under both expected and unexpected conditions. They achieve this by compensating for fluctuations, disturbances, and environmental changes in a situational manner. Approaches to increase resilience are referred to as “resilience engineering”. Within this framework, own methods are developed (e.g. Hollnagel et al. 2014) or appropriate methods from other fields are applied, such as planning games and educational games (e.g. Schuh et al. 2020), that help to generate and use mental models.

The resilient system is not as much of a methodological tool in contrast to previous system concepts, it is rather a guiding principle or a model for the development of systems pertaining to the conditions of complexity. Modelling and developing systems can be utilized through theories on factors of resilience (Weick and Sutcliffe 2015; Hollnagel 2018; WHO 2022). In this way, people in the system, at the system and far away from the system can communicate and develop with each other.

In order to meet the requirements of a digitalized world of work with its complex socio-technical systems, the integrative approach presented must be applied. One way to do this can be through industrial and organisational psychology programmes at universities of applied sciences (Mühlbradt 2016), if the universities succeed in strengthening the field of “work and technology” and teach competences needed for dealing with complexity. The following section outlines how this approach is implemented in the training of industrial and organisational psychologists at the FOM University of Applied Sciences in Aachen.

6 Digital Work Design at the FOM Aachen

The FOM University of Applied Science, based in Essen, North Rhine-Westphalia, is a non-profit private educational institution. It is currently the largest (nonvirtual) university in Germany with over 30 locations in Germany and Austria and more than 57,000 enrolled students. Most students have a vocational qualification and initial work experience, they work and therefore study at the university in part-time.

The industrial and organisational psychology department is the second largest department at FOM university with around 10,000 students. This number represents a share of about 10% of all psychology majors in Germany. The Bachelor of Science (7 semester, 180 ECTS credits) and Master of Science (4 semester, 120 ECTS credits) degrees are offered for industrial and organisational psychology at the FOM university. The educational goals of these programmes are formulated as competencies and are further subdivided into professional, methodological, social, and personal competencies. The modules include business administration and psychological content. The Master’s programme includes the following modules (Table 1).

Table 1. Programme and modules of the master of science in industrial and organisational psychology with credit points (CP)
Full size table

Around 600 students are enrolled at the FOM university in Aachen. The location is characterized, among other things, by its proximity to the large state university (RWTH Aachen) with which joint research projects are carried out. The digitalization of work is one focus of the Aachen location. Within the Master’s degree, the study content and forms of work and organisational psychology components are increasingly oriented towards digital work design in a world with complex socio-technical systems. In doing so, references to contemporary research are made. Following, this will be considered in more detail for some modules.

Qualitative Research Methods

There is a remarkable parallel between the axioms presented above and the postulates of Mayring (2016), which he understands as methodological principles of qualitative research in distinction to quantitatively oriented research. Qualitative thinking relates to “systemic thinking”, because it also assumes complexity, which it does not want to artificially simplify. This clarifies the meaning of the integrative approach once again in concrete terms for the person, the endeavour of the work analyst and work designer, because Maying’s postulates also contain direct demands on the acting person. The signals that Mayring referred to as “...qualitative turn ... [as] ... trend towards qualitative methods of cognition...” (Mayring 1986, p. 9) find parallels to Braithwaite et al. (2017, p. viii) for example, who distinguished between “complexity thinking” as opposed to “linear thinking”.

In the above-mentioned study programme, qualitative methods are on an equal basis with quantitative methods. Therefore, theoretical basics, methods and instruments are taught. Where complexity cannot be resolved, qualitative methods provide suitable approaches to data collection and evaluation. The module has a high proportion of exercises on the methods and techniques used that are taught in theory and practice, especially considering the integrative approach, the methodology of the Functional Resonance Analysis Method (FRAM, Hollnagel et al. 2014). The FRAM can be attributed to resilience engineering while the methodology is essentially based on qualitative interviews with job holders. The aim is to show actual processes in everyday work and their interconnectedness. The interview results can be transferred into a visual model of the process, for which a special software (FRAM visualizer) is used. Results from ongoing research are incorporated into the lectures as well as experiences regarding the application of the FRAM method (Unger et al. 2022).

The combination of quantitative and qualitative methods and the elaboration of their particular characteristics as well as knowledge regarding advantages and disadvantages offers students a broad selection of methods for analyses and controlled interventions.

Industrial and Organisational Psychology

The module industrial and organisational psychology (e.g. Nerdinger et al. 2019) is deliberately focused on the aspect of “work and technology”. The lecturer discusses references and current research as well as recent developments on the subject (e.g. Adolph et al. 2021; Dworschak et al. 2021; Schuh et al. 2021; Mühlbradt 2022). Table 2 displays an overview of the module’s topics.

Table 2. Topics of module “Industrial a. Organisational Psychology” Summer Term 2022
Full size table

Also, topics are taught that cannot be exclusively assigned to the psychological discipline. These topics range from information about other disciplines that shape work, to the basics of automation and machine learning as well as platform economics. Therefore, psychological content must yield sometimes due to limited time resources. In doing so, we follow the view of Rosenstiel (2004, p. 89), that, depending on the specific field of work, non-psychological knowledge can be more relevant than psychological knowledge.

The module industrial and organisational psychology is linked to the module Qualitative Research Methods, as it also provides in-depth information on the FRAM. In terms of methodology, the module focuses on discovery and cooperative learning in small study groups. A script serves as guideline that includes selected original works and case studies. The script is structured into topics with leading questions to each topic formulated for study groups. The lecturer’s role is that of a coach after a short presentation of the topic, the materials, and the leading questions. Finally, the study groups present their results and discuss them in the plenary.

Elective Module Independent Research Project

Students can choose between marketing psychology and industrial and organisational psychology in the elective module of the third semester. The following explanations refer to the second option. Students work alone or in teams of two to accomplish an industrial organisational psychology analysis or develop a concept for a company or an institution over the course of the semester. The relatively long time period enables the students to work intensively on a topic and at the same time provides the opportunity to conduct data in the field while the lecturer accompanies and coaches the students during this process. In order to gain practical knowledge as a student it is a good concept to approach participating companies during an ongoing collaborative research project of the university, especially if located in the region. This way, the lecturer ensures that the resulting assignments for the students are demanding, but not overly extensive. Examples of assignments in this module are:

  • Risk assessment of mental stress in companies: Consulting and development of appropriate methods, implementation, and evaluation of surveys

  • Competence modelling and measurement in companies as well as coaching for employee evaluation and development meetings

  • Conception of the unit “Leadership in the digital world of work” within the context of a training concept for companies

Looking at it from a professional perspective, Master’s students are able to meet the requirements in the third semester. They also benefit from their life and professional experience within their studies. The coaching of the lecturer therefore includes topics such as “clarifying assignments”, “planning work and time” and “how to deal with obstacles”. These are important aspects of self-management.

Psychological Competence and Transfer Assessment

Complexity may cause ambiguity, uncertainty, and conflict when linear approaches fail or incompatible views or models of reality clash. It can be important to support people in or at the system to question, change and exchange their models of reality based on the credo “develop instead of design”. Skills are therefore taught and trained through case studies, role plays and exercises in the module Psychological Competence that address these topics:

  • Perception and reality

  • Active listening and solution-oriented conversational skills

  • Dealing with criticism and conflict management

  • Crises as starting points

  • Coaching

In the Transfer Assessment (TA), students are encouraged to independently compare their level of competence and their competence expectations and thus actively reflect on their competence development. In addition to the kick-off event for the TA as part of the decision-oriented management module in the first semester, events for the presentation of the transfer reports and for transfer reflection take place in the second, third and fourth semesters. Two transfer questionnaires, three transfer reports and a part of the oral examination at the end of the degree programme are required for the examination. The transfer assessment promotes self-reflexivity and runs over four semesters. Expectations of competence development and self-perceived changes with regards to competences are explicitly formulated and documented.

This way, the written and detailed reflection of selected modules facilitates the students own learning and provides an increase of awareness about change processes. These experiences can help to design change processes for other people in the future.

7 Epilogue

In this paper, an attempt has been made to show how the rise of digital work contributes significantly to the dominance of complex socio-technical systems as an essential organisational form of work and value creation. This leads to a crisis of work design, as existing complexity theories and approaches derived from them are ultimately inadequate. In contrast, the integrative model demands the merging of two previously unconnected strands of systems theory so that systemic thinking becomes an integral part of the analysis and design of socio-technical systems. The central idea is that complexity is the same for all actors, cannot be resolved and is ultimately indispensable for achieving the goals of modern work systems. Subsequently, the consequences of this perspective for tasks of and demands on labour analysts and designers were considered.

Healthcare, being considered the laboratory of complexity, Verhagen et al. (2020) indicates that the obstacles in the practical implementation and dissemination of systemic approaches to patient safety that include fundamental changes, do not happen overnight and without problems. In line with this, St.Pierre et al. (2022) still find a too narrow thematic focus on traditional approaches to risk and error management considering the twentieth anniversary of “To Err Is Human”.

The legend of the Gordian Knot, which no one can untangle, is well known. Nevertheless, whoever manages it once shall rule. Alexander the Great solved the problem by ignoring the goal of a non-destructive solution to the problem and was thus able to simplify his task considerably: He cut the knot with his sword. The myth of Alexander the Great is still powerful as a metaphor for resolving complexity. However, the appropriate way to deal with complexity is to recognize it as a borderline experience first. This may well be experienced as a loss of control. However, in the digitalized world of work with unleashed work and high demands on the performance of systems, it will be difficult to find an alternative way in the long run.