1 Introduction

In this paper we aim to contribute to the on-going discussion of the characteristics of a joint system design approach. The design concept by Woods and Hollnagel (2006) is a central reference in the discussion, but we shall also draw on other concepts, e.g. those proposed by Elm et al. (2003) and by Albrechtsen et al. (2001). Our contribution deals especially with three points: what are the design pressures that would demand a change in the cognitive systems engineering (CSE) paradigm just now? How do we (in empirical research and design) define what elements belong to the functionally united joint system? What challenges does the joint system approach put on the theory of human conduct? Via discussing these issues we shall outline what we see to be the logical phases of a design process that is oriented to create an appropriate joint system.

Human factors is a science of design that exploits psychology and other human-related disciplines in shaping technology to meet the needs of the human user. Yet, it is not self-evident how human factors could be made effective in the design process. Depending on the perspective taken to the human actor, different targets become dominant in the design agenda. At least three basic assumptions of the role of the human with regard to technology may be distinguished: human as a risk factor, which approach leads to minimising the negative effects of the human actor and to building as comprehensive automation as possible; human as a creative factor, which notion facilitates design to maximise the benefits of the user and to develop technological support to him/her; Finally, human conceived in a co-evolution with technology. In the last case design aims at creating new possibilities and uses of living (Hancock and Chignell 1995; Papin 2002), i.e. it aims at design of practices.

Different theoretical threads in psychology support the above mentioned design rationales. The “information-processing theories” match with intentions to cope with human errors. The “cognitive oriented mental model theories” identify differences in human cognition and provide incentives to improve the creative role of the human via design solutions. In this theoretical wave also the notion of CSE was invented. It was a new approach to human factors design that took a critical position to the information processing approach (Hollnagel and Woods 1983). The CSE approach facilitated the attempts of the 1980s to design decision-support based on artificial intelligence. It responded well to the needs of the human–computer interaction design and won a strong position in explaining human cognition and action.

Both the information-processing and CSE approaches have maintained their position notwithstanding evident drawbacks. The critique towards these approaches defines what is expected of the third human–technology interaction theory trend, the “systemic approach”. It may be labelled as an ecological approach as it focuses on transactions between the different elements of the system. The critique expressed from the systems perspective first focused on the deficiency of information processing approaches to consider the contextual dependence of cognition and action. This issue was discussed e.g. by Suchman (1987) as she advocated the situated organisation of action. Later Hutchins (1995) took distance from focusing the analysis on an individual user and demonstrated how to describe cognition as being distributed within a group of people, technologies, and even over time in the form of cultural artefacts. Dourish (2001) drew attention to the embodied nature of human perception and action and described the tangible and social nature of the computations that the technologies support. Both Bannon (2002) and more recently also Norman (2006) pointed out the weaknesses of the concept of action. Action-oriented design is focused on an individual user and on his immediate tasks and experiences without considering the broader context of conducting. Hence, these authors propose that the more holistic concept of activity is more pertinent in the design of technology.

Hollnagel and Woods (2005, pp. 18–19) propose a strategy for identifying what is the common cause of the current theoretical difficulties in CSE. According to the authors it is the disintegrated conception of the human–machine relationship, which leads to postulating an “interface” that connects the two separate elements. Adopting an integrated approach would “change the emphasis from the interaction between humans and machines to human–technology coagency” (p. 19). The human–machine coagency should be understood as a functional unity, a joint cognitive system (JCS): “JCS is not defined by what it is but what is does” (p. 22, italics in original). The authors see that the shift in perspective denotes a change in the research paradigm regarding human–technology interaction.

Indeed, the proposed solution of considering the human–technology coagency is plausible. It may be connected to the claim for analysis of activity that e.g. Norman (2006) has expressed, and it finds support from other authors, too. The idea of functional unity of diverse cognitive elements characterises the distributed cognition approach (Hollan et al. 2000; Hutchins 1995). Phenomenology-oriented theories of embodied mind (Lakoff and Johnson 1999), ecological psychology (Gibson 1979), and human–environment system theories (Ingold 2000; Järvilehto 1998a) also draw on the idea of functionally united systems. Also we lean towards the idea of functionally united joint systems (Norros 2004, pp. 28–30).

2 Current design pressures: flywheel for a paradigm shift in cognitive systems engineering

As was indicated in the previous section, the idea that human and his environment—including the technical tools—constitute a functional unit is as such not a new idea, but yet, this way of thinking has not become dominant. Now, Hollnagel and Woods argue for a presently on-going paradigm shift in the CSE and make a reference to Thomas Kuhn (Hollnagel and Woods 2005; Kuhn 1970). According to Kuhnian thinking, verifying that such a radical change in theoretical thinking really is taking place requires, first, that an explicitly formulated alternative should be available. In this case it is the joint system approach. Second, reasons to question the relevance of problems solvable within the current paradigm should also be evident, or doubts should be expressed concerning the interestingness and meaningfulness of the solutions that the presently dominant information processing paradigm offers. Demonstrating such tendencies is not easy, because the facts acquired within the old paradigm cannot be proved false by the alternative paradigm. The latter will have to identify relevant new problems and produce convincing empirical facts and good design solutions concerning them.

Our own experience is that it takes many years to comprehend, and then to demonstrate what the new problems are, with regard to which the joint system approach will show its strengths. In our case, the theoretical interest in the joint system approach emerged over the years in studies on operator work and decision making in different process control domains. Yet, not before a recent involvement in outlining a design concept for future “intelligent environments” did we comprehend novel design constraints and challenging human–technology interaction problems that are out of the reach of the old design paradigm and its theoretical background concepts.

A new setting was also required for the research. A multidisciplinary project group was formed that gathered members from diverse theoretical orientations. It included both researchers and designers. The group discussed frequently with an industrial reference group that was set up of people from firms. During these discussions the need for conceptual and methodical work was identified as a prerequisite for innovations. The project group agreed that its aims are to outline a new design approach that should be based on “ecological systems thinking” and to focus on the “design of intelligent environments”. The project articulated reasons for claims for paradigm shift in CSE and also outlined new problems to be solved. In the following we shall present the main chain of argument developed within the project (Kaasinen and Norros 2007; Norros 2006). Our argument for the need for a new design approach consists of five steps (see Sects. 2.1 to 2.5).

2.1 The object of design changes from product to “intelligent environments”

We start by stating that information and communication technology (ICT) creates new possibilities to human usage due to massively increasing computational power, and due to ICT’s capability to expand context in which its power may be put into use (Dourish 2001, p. 2; Rückriem 2003). Hence, instead of being used in single equipment-like products technology becomes ubiquitous and a part of a new type of “intelligent environment”. As technology is embedded in the environment it shapes what we have considered natural environment. In this context, “intelligence” is not an attribute of technology or the human element as such but, instead, it refers to the appropriate functioning and adaptation of a system. The whole adaptive system, or “intelligent environment”, becomes the object of design.

2.2 The new object brings design tensions

The second phase of the argument is that intelligent environments put inherently contradictory constraints on design. In the discussions during the project work we identified phenomena that would characterise situations that the designers face. We identified a group of “design tensions” that characterise the dynamics of the design field on a general level. These dilemmatic situations demonstrate that intelligent environments put both broader and deeper problems for cognitive systems design than those with which the designers are used to deal with. The design tensions are indicated in the following Table 1 (Keinonen 2007; Kuutti et al. 2007).

Table 1 Trends in the design of intelligent environments and the tensions they create to design (with terminological adaptations Keinonen 2007)
Full size table

The first two tensions of Table 1 refer to different views of defining the design object and the pressures that arise from there. “Technology tension” emerges between the vision of intelligent environment characterized by extensive use of ICT in all spheres of life, on the one side, and the actual everyday needs of people that may be fulfilled intelligently by a minimum usage of technical support, on the other. The balance must be found by developing a design process that does not exclude either technology or human practices. “Innovation tension” refers to different views of the temporal structure of the product, i.e. is the product a distinctly novel entity, or something that evolves step by step. Hence, balancing must take place between radical and rapid changes, on the one side, and piecemeal construction upon already existing solutions, on the other. Business interests and management perspectives may emphasize the first alternative while other values like safety or sustainability considerations would prefer the second.

Three further design tensions arise from the design process and practice. “Competence tension” refers to the diversity of competences that are needed in the design of intelligent environments. By “user designers” we denote users that contribute to the design process by understanding the usage and future needs from inside the practice. Design requires, however, also scientific and special knowledge, design skills, and mastery of whole design processes. Therefore “researcher designers” are needed. Jointly mastered vocabulary, models, and transparency of the course of design process are required for balancing between these demands. “Readiness tension” focuses on the dilemma between the needs and expectations to take the product into use immediately and without effort, and those to proceed stepwise, combine partial solutions and to tailor solutions in practice. Finally, the “generality tension” draws attention to the fact that due to the embedded nature and connectedness of the different elements of the intelligent environment, and to the need for making practical experience useful in design, generalisation and conceptual knowledge is needed to steer design and to reach more distant aims. At the same time, focusing on specific design solutions is immediately motivated, and economically pressing.

Many of the above presented challenges have existed and influenced design already for long in some form or magnitude, and the flexibility of design discipline has digested changes. Articulating the trends, however, increased understanding of the constraints and possibilities that characterise the present and future design situation.

2.3 Coping with tensions creates new design activity

The third step in the chain of argument is that coping with the above mentioned design tensions will lead to changes in the design activity and its processes. Tensions create constraints and practical design demands that the new paradigms of design and CSE will have to face. New solvable problems will emerge but they must be conceptually grasped. On a global level, two tendencies could be distinguished in the analysis of the design tensions: one is the strong focus on the users and their local practices; the other is an equally demanding tendency for distancing from the immediate design situation and solutions, and for putting increasing attention on generic issues and preconditions for design. Both tendencies have consequences on the present design practice and the qualifications of the designers.

It is proposed that while “product design” as the core practice of design will remain relevant, two complementary design modes emerge. These are labelled “immediate design” and “remote design”. Characteristic to immediate design is its sensitiveness to the users’ current or expressed future needs, its context dependency and intensive utilisation of layman designers. Immediacy of this design mode does not only refer to time and location but also to the causal and value-based immediacy, i.e. the needs of the users are the immediate reasons for design, not e.g. business strategy or technical opportunity (Keinonen 2007, p. 10). Remote design, again, aims at structural changes, focuses on possibilities and is typically formative i.e. outlines general solutions. Scientific work and innovations are needed to accomplish results, impacts of which become materialised later e.g. when technologies have matured sufficiently to application. The context of implementation of the results of remote design expands beyond the original area of application, and design is motivated by societal and technological policy reasons.

2.4 Characteristics of design knowledge

In the fourth step of our argument we reason that the nature of the “intelligent environment” and the completion of the modes of design (from product design to immediate and remote designs) are the frames within which new design problems will arise. These will be solvable but require an elaborated understanding of what design knowledge is like (Norros et al. 2007). We looked into science research, design research, human–technology interaction research, and innovation research to find seeds for what we called “ecological systems thinking” in design. An important criterion for defining such a way of thinking is, first, how it understands the nature of knowledge and knowledge production in design. Second criterion that we consider important is that the approach is capable of handling human and technical elements of the intelligent environment in transaction with each other.

Regarding the epistemological basis of the ecological systems thinking in design we are attracted especially to the ideas of Gibbons et al. (1994). These authors make an important clarification of the nature of knowledge needed for design and how it is produced. Gibbons and collaborators claim that the traditional model of production of knowledge is close to what is meant by academic knowledge production, i.e. science. This mode of knowledge production is labelled “Mode 1” by the authors. It is disciplinary organised and hierarchical and its quality is determined within the scientific community. A complementary mode of knowledge production is also emerging. It is labelled “Mode 2”. According to the authors it is different from Mode 1 almost in every respect. Mode 2 is created in the context of its application, it crosses disciplinary borders and makes use of concepts that enable focusing on actual and particular problems, it acknowledges divergent contexts of knowledge production from science laboratory to hands-on experience in the field, it is reflexive or dialogic rather than strictly objective, and finally the quality of Mode 2 knowledge is evaluated broadly including its practical relevance (Kuutti 2007; Kuutti et al. 2007). Acknowledgement of the validity of both Modes 1 and 2 types of knowledge for design is necessary if we are to fulfil the knowledge demands of all the design modes, i.e. “product design”, “immediate design” and “remote design”. Equally important from the design point of view is that Modes 1 and 2 are not considered competing but instead complementary forms of knowledge production, which can be combined in an actual design cycle.

2.5 New concepts to tackle the design object

Finally, Gibbons’s notion of design knowledge guides the definition of new concepts that would enable tackling and designing human and technical elements in transaction with each other. The methodological principle that human and environment form a functional unity is one of the important new ideas. This demand has been identified in our applied research on complex high-reliability work, which has lead to formulation of a contextual methodology for work analysis labelled the core-task analysis (Norros 2004).

In the present connection, as we are especially interested in design activity the object of which is the intelligent environment, the human–environment system notion must be developed further. It is especially necessary to elaborate on the relationship of human and technology within the joint system frame. Furthermore, the joint system must be tackled on empirical level. With these intentions three approaches to joint systems will be compared and discussed in the next chapter.

3 How to define what elements belong to the functionally united joint system?

So far we have discussed the broader reasons for and benefits of adopting a joint system approach for CSE. This solution appears to provide the possibility to understand intelligence and functioning of a system consisting of different kinds of elements. But we have not yet considered precisely what elements constitute a functional unity and how to define empirically the elements. It would also be fruitful to clarify how different approaches of CSE that basically agree on the need for a joint system approach, understand the system. Dealing with these questions made us aware of a further one: is the joint system to be understood as a unit of analysis or a concrete object of design, or both?

In the following the above questions concerning the nature of joint systems will be discussed with regard to three approaches that clearly express the intention to re-define the relationship of human and technology. These approaches are the JCS by Hollnagel and Woods (2005) and Woods and Hollnagel (2006), an extended CSE approach proposed by researchers within the Centre for Human–Machine Interaction (CHMI) and at Risö National Laboratory (Albrechtsen et al. 2001), and the joint intelligent systems (JIS) approach proposed by the present authors at VTT Technical Research Centre of Finland. These approaches share a common background in the CSE of the 1980’s and have been influenced by the ideas of Jens Rasmussen from Risö (Rasmussen 1986).

3.1 Joint cognitive systems: the approach of Hollnagel and Woods

The JCS approach proposed by Hollnagel and Woods (2005) and Woods and Hollnagel (2006) has been published in two recent books. The work has already raised great interest and applications are emerging. The authors use the term “coagency” to denote the idea that the focus of analysis is not human–technology interaction but the functioning of the human–technology–work system as a whole (Hollnagel and Woods 2005, e.g. p. 19 and 67). Regarding the functioning of the system, the authors find especially interesting to discover how a mutual “adaptation” of these elements takes place (Woods and Hollnagel 2006, p. 7). In this connection the authors draw on Uexküll (1934) and Gibson (1979). These classical authors claim that understanding adaptation assumes that each element of the system is seen to be defined by the other elements and that adaptation always requires changes in each element. As a consequence “demands” emerge on the functioning of the system, and strategies are created to maintain its functioning. Woods and Hollnagel theorize further that three types of adaptive mechanisms characterise the functioning of the system. These are labelled patterns. Patterns are more or less ostensively defined, i.e. by examples of different patterns. The three forms of patterns are:

  1. •

    Patterns of coordinated activity—how cognitive work is distributed and synchronized in the changing world

  2. •

    Patterns of resilience—how anticipation and adaptation to potential surprises take place

  3. •

    Patterns in affordance—how artefacts support people in the face of the demands of work.

When discussing the patterns of affordance the authors deal specifically with the human relationship with technologies. It becomes evident that in the JCS approach affordance is considered a relationship between the human actor and the artefact, but that the environment is not included as part of a joint cognitive system. This definition of the joint system escaped our notice first, maybe due to our strong mindset of a unity of human and environment, rather than human–technology, but also because in many places the authors refer to work setting when talking about the human–technology system (Woods and Hollnagel 2006, e.g. p. 7). For example, as the authors define the system goal-oriented (Hollnagel and Woods 2005, e.g. p. 115) the interpretation emerges that goal refers to environment, and, further, that environment would be part of the system. It remains that it is not immediately clear whether and how the real world, in which the system is located and that becomes an object for intentions, is included in the system, or not.

The authors prefer to utilise generic system theoretical vocabulary to characterise the joint cognitive system. In this context “environment” is something that by definition does not belong to the system. The system border is something that has to be defined, and this definition is based on the functions the system should serve (Hollnagel and Woods 2005, p. 45). The border of the system is not considered absolute, but relative to the aims of the analyst, i.e. the boundary “depends on the purpose of the analysis” (Hollnagel and Woods 2005, pp. 115–116), not on the activity and the reasons of people who are using technology and whose actions the analysis is about.

From the viewpoint of design, patterns are empirical generalisations abstracted from studies in specific situations and connections. Generalisation and abstracting of patterns enable the analyst to identify ways of system adaptation. Because the focus is on explaining global dynamics of joint cognitive systems it is not considered in detail how the generic patterns of adaptation are manifested in the activity of the actors themselves. For example we did not find considerations of how to analyse affordances on empirical studies in situ, or how users, for their part, may become part of the joint system patterns.

Expectations that the user’s point of view would play a role in defining the functioning of the system are raised because the approach is characterised as “practice-centred” (Woods and Hollnagel 2006, p. 5). In this connection the phenomenological approach could be useful, and it is referred quite lengthy by the authors (2005, pp. 93–99). It remains open, however, what role the experiential and user-driven point of view to the human–technology–environment relationship really plays in the analysis of the dynamics and generic patterns of joint cognitive systems.

The joint cognitive system is clearly proposed to be the new unit of analysis in CSE. In coherence with this, an interesting process model to define different aspects of knowledge production is labelled as a “practice-centred approach to research” (Woods and Hollnagel 2005, p. 5). Whether the joint cognitive system should also be interpreted as the design object is not clearly expressed but we interpret it to be the case.

The JCS approach makes sense to us. We consider that the notion of pattern is central in the approach because just this concept enables drawing attention to the functioning of the system, instead to its structure. We feel, however, that it would be necessary to define methodologically what a “pattern” is, and why it should serve well in defining the system’s way of functioning. It appears to us that, by asking how using the notion “pattern” differs from using the notion “action” in describing system functioning, would clarify the specific features of a “pattern”.

3.2 Joint intelligent system: our approach

The joint system approach that has been developed by us at VTT has been labelled the JIS approach. This name was chosen within an interdisciplinary VTT project that focused on developing a new design approach under the heading intelligence engineering framework (IEF). The human factors partners of the group insisted that intelligence engineering should consider intelligence as an attribute of a joint human–technology system. The idea found support in the project and it was agreed that the products to be seen as objects of intelligence engineering should have the characteristic of a JIS. One of references used to argue for the joint system was clearly the work of Hollnagel and Woods (2005) and Woods and Hollnagel (2006).

As indicated above, the notion of “pattern” stroke us in the JCS. It resembled the ideas that we had also been playing around in our earlier work. Patterns are, as discussed in the previous section, used in the JCS approach to describe joint system functioning. In this connection we take the opportunity to elaborate on this issue of patterns. But before we proceed to this central topic, some background for our JIS concept.

Our attempts to re-define the relationship of human and technology in complex work systems owe much to a Finnish scientist Timo Järvilehto. His research area is general psychology and brain research, and he has studied human perceptual processes, especially tactile and cutaneous senses. Järvilehto (1998a) has also contributed to methodological and philosophical questions of the relationship of human and environment. He advocates the basic premise that from a functional point of view, human (or other organism) and his (its) environment form a unity. “Environment” denotes that part of the world that may potentially be useful for a particular organism. Among even more classical authors, Järvilehto draws on Gibson (1979) and Uexküll (1934), as do Hollnagel and Woods in their earlier cited work. Järvilehto (1998a, b, 1999, 2000) has presented detailed evidence to support this position and demonstrated that adopting joint system approach has consequences on how different psychological phenomena are comprehended.

Important to our present discussion is the idea that aimed results of the activity, in which people are involved, determine the content and structuring of the emerging human–environment system. Hence, various forms of human, technical and other environmental elements become attached to the system depending on what is needed to maintain an appropriate functioning of the system, i.e. activity. Hereby we come to define the basic structure of the joint system: it is a system comprising human, technology and environment. The functioning of the system is organised according to the results that are aimed at. In other words, the system is organised by its purpose, and shaped by the constraints and possibilities of the environment which must be taken into account to maintain adaptive behaviour.

The above characterisations are in accordance with the basic notions of CSE (Rasmussen 1997; Vicente 1999). Within CSE there exist several methodologies that define how to approach complex system behaviour on empirical level. Some examples are the cognitive work analysis (Vicente 1999), or cognitive task design (Hollnagel 2003a), and applied cognitive work analysis (Elm et al. 2003). These approaches are methods that guide the analysis of the behaviour of a particular human–technology–system in actual situations. Also the naturalistic decision making literature provides a number of interesting approaches to describing actual behaviour of the system (Montgomery et al. 2005; Zambok 1997). Accumulation of results of such analyses surely provides us information of the frequency of different behavioural phenomena and their appearance with regard to different conditional variables.

But the point is that methods described above do not inform us of how systems behave. By this we mean that these analyses do not explain the way of working, i.e. patterns of behaviour, to use the notion that Woods and Hollnagel (2006) prefer. “Way of working” is not a generalisation of externally observed behaviour but, instead, an expression of an internal regularity in the behaviour of the system. The latter is what we try to capture by exploiting the semiotic concept of habit (Peirce 1998a) and use it in empirical analysis of people’s usages of their tools and technology (Norros 2004). These usages are examples of joint system functioning.

Habits, in pragmatic terminology, are dispositions to act that the system (in our case the joint system) has acquired when dealing with different environments and situations (Dewey 1997; Kestenbaum 1977; Peirce 1998a). Habit provides an interpretative i.e. generalised internal potential to act that enables action in ever changing environments. Without interpretation action tends to become reactive as Peirce maintains (Peirce 1998b). Habits are not just results of repetition, but, at the same time they are repeated because they express what is adequate or meaningful in a particular situation, and hence worth repeating.

What, then, would be the role of tools and technologies in habit? Here it is instructive, first, to analyse how human action is conceptualised in the cultural historical activity theory (CHAT) the founder of which was among the first psychologists to analyse the role of tools in human behaviour (Vygotsky 1978). Vygotsky’s conception of action is depicted on the left-hand side of Fig. 1. It states that human actors are related to the environment (object) in a mediated way via instruments or signs. Vygotsky is known for his important discovery that beyond the instrumental influence that a tool has on the environment, called the instrumental function, an instrument also has a psychological function. The latter refers to the principle that a sign or instrument to be used in action creates and assumes a schema of its use by the actor.

Fig. 1
figure 1

On the left-hand side the structure of mediated action in the cultural historical activity theory (CHAT) (Vygotsky 1978) is depicted. The instrumental and psychological function of an instrument or sign may be described by this model. The dashed line indicates that connection between actor and object is typically not direct but mediated. On the right-hand hand side can be found the semiotic model of action proposed by Peirce (1958). Signs/instruments are connected to their object through the mediation of the interpretant. This model portrays habit and is capable of describing the communicative function of instruments and signs

Full size image

Pragmatist thinking is in agreement with the above described but would maintain, however, that explication of action in the way CHAT or other psychological theories of action do, refers to singular situation specific actualisations of human behaviour. What is needed, therefore, is to understand the meaning of behaviour that the mediated structure also portrays. Explicating the semiotic aspect or instrument usage is depicted in Fig. 1 on the right-hand side. It portrays habit. A similar triangular model is used but instead of actor the notion of interpretant is used. The figure states that instruments or signs are related to an object or purpose by a behavioural form that is an interpretation of the connection (interpretant). This behavioural effect expresses what the sign means in the given context.

The semiotic structure completes the analysis of action by defining habit. In other words, instruments and signs, by being connected to objects over behavioural effects, also convey the meaning of using them in that way in a certain context. By acting in certain ways, including possible communication concerning acting, people do not just act more or less appropriately with the tools (instrumental and psychological functions) but they also mediate what they consider for relevant and worthy. Due to this, the sense of acting is mediated to co-actors and the behaviour of people becomes predictable in the larger community (Norros 2004). Hereby a third, communicative function has become overt. This function was not originally considered as a separate function in CHAT. Recently at least Rückriem (2003) and Bödker and Andersen (2005) have drawn attention to the need to articulate the communicative function, and these authors discuss its position in the analysis of action within CHAT.

Our conclusion is that in order to capture habits, or patterns of joint system, it is necessary to be sensitive to all three functions of tools and technology. Considering the instrumental and psychological functions is necessary for the analysis of situational and particular actions. In order to reach the generic interpretative aspect of action it becomes necessary to elaborate also how tools function as communication media. Explicating the communicative function enables comprehension of the meaning of the actions and tools in the community. Revealing the meaning is the central aspect of understanding what is general in action, because it is exactly the internal reason why the structure is repeated. Habits are practice internal regularities or patterns. The connection between action and habit and the three functions of tools may be depicted by using the model proposed by Bödker and Andersen (2005, p. 363). This extended model of action maintains that the CHAT model of action and a semiotic model of communication can be related by assuming that object, on the one hand, and signs and instruments, on the other, may be considered simultaneously as parts of an instrumental and communicative action. This assumption is in accordance with our thoughts of the connection of action and habit, and the three functions of signs/instruments. Our adaptation of the model is depicted in Fig. 2.

Fig. 2
figure 2

Action and habit may be conceptualised as parts of an extended model of mediated action. The instrumental, psychological and communication functions of signs and instruments may be analysed with this model. Adapted from Bödker and Andersen (2005, p. 363)

Full size image

The identification of habits in empiric analysis requires that human actors have a chance to express their reflected or non-reflected reasons. We are interested in reasons and purposes relevant to those activities, in which people are involved in their work. Work activity provides the context out of which reasons emerge. Therefore, we have developed a modelling technique within the core-task analysis approach (Norros 2004; Savioja and Norros 2008) that combines the CHAT-based analysis of core-task demands and the functional domain analysis proposed by Vicente (1999). The role of this modelling is to enable understanding of the possible reasons to act people may have in the activity in question. Efficient reasons are inferred from observations and analyses of actual acting.

In our case, the joint system functioning is defined as an internal regularity of the behaviour of the system. The system comprises functionally relevant parts from the environment, human and technology. Emergence of such internal regularities requires communication within the system and reconfiguring of the three main elements of the system (human, technology, environment) according to the changing demands. Communication assumes that each system makes itself affordable by the other. Design should be aimed at enabling communication, reconfiguration and affordances. These demands characterise the design of a JIS that, for its part, could develop internal patterns of functioning to cope with the changing demands put on the system. The demands on design are close to those that Hollnagel and Woods (2005) describe as patterns. These patterns are, indeed, to be interpreted as design patterns of cognitive engineering. We agree that new concepts are needed for designing adaptive usages of tools, i.e. adaptive practices of joint systems. “Design for adaptation” was highlighted also by Vicente (1999) as the desired attribute to design.

3.3 Design of affordance spaces: the Risö-CHMI approach

New ideas to CSE have also been expressed by researchers within the Centre for Human–Machine Interaction (CHMI), and at Risö National Laboratory (Albrechtsen et al. 2001). The group includes members from the human computer interaction field, and others who represent the long tradition in CSE at Risö. Due to its orientation towards real contexts of human performance and the focus on extensive and complex technical systems the latter approach is valued high and it has a central place in the development of the new approach.

Theoretically the approach leans strongly on the ecological psychology of Gibson (1979). The central premises of this theory are the immediate perception of the possibilities, i.e. affordances provided in the environment for human use, and the idea that affordances are not strictly objective features of the environment. These original notions are emphasized by the authors. Hence, the idea of functional systems between human and the environment are present in the Danish approach even though the issues of one or two systems is not articulated—at least not in the referred report that otherwise includes an elaborated theoretical account.

The current ways of making use of the Gibsonian theory in human–computer interaction research are found problematic and too narrow by Albrechtsen et al. (2001). As the point of departure in their development of an elaborated concept of affordance the authors make two critical points. First, they see, that typical of the present interpretation of affordances in human–system interface design is that affordances are considered as add-on surface phenomena of artefacts (p. 31). They admit that the strength of this interpretation is that it provides a possibility to deal with generalized features of artefacts and allow universal treatment of artefacts over various domains. The drawback is that the artefacts are considered detached of the context of actual use. Secondly, the authors state that one of the most influential usages of the notion of affordance, i.e. that proposed by Donald Norman, implies a reductionist perspective on affordances. According to this, affordances are conceived as static features or objects of the world/system etc. rather than as dynamic and changing properties of the environment affording action (p. 31). The authors state, that even though this interpretation may be plausible based on Gibson’s original ideas of affordance, he did not define affordances as inherent features of the environment nor the actor. Instead they should be understood as dynamic elements evolving through the situational coupling between the actor and the environment (p. 31).

As a possibility to extend the affordance concept the authors turn to the cultural-historical activity theory (CHAT) (Leont’ev 1978; Vygotsky 1978) and CSE (Rasmussen 1986). It is seen that these provide a frame to consider affordances as constituents of human work activity. The authors take the socio-historical, cultural and organisational dimensions, goals and constraints as important aspects in defining affordances. The programme the research group has set itself is to develop a synthesis of these three main sources of their theoretical work, Gibsonian approach, CSE and CHAT. It is made clear that the primary unit of analysis is the human work activity and the socio-cultural context in which this activity is carried out. Hence, a shift of the unit of analysis from action to activity characterises the new approach.

We interpret that even though Albrechtsen et al. (2001) define the unit of analysis to be activity, the focus of design efforts is connected to developing affordances. From the design point of view it is very interesting that the authors make an attempt to consider also the meaning or relevance of the environment to the actor, i.e. to develop a semiotic interpretation of the concept of affordance. In considering this the authors refer to the work of Pickering (1999) who draws on Peirce’s semiotic theory. Pickering sees that affordances could be regarded as types of signs to which the organism adapts its actions in the course of its evolution. “In Peirce’s terms, affordance is a sign for which the organism acts as interpretant to produce action in a given situation as the object. Thus organisms do not merely respond to stimuli, but act on the basis of meaning” (Pickering 1999). The semiotic point of view—that has a strong connection to our own analysis—is not further developed in the paper by Albrechtsen et al. in the discussed report. The possibilities of this perspective are, however, noted as one of the two major challenges for further work (Albrechtsen et al. 2001, p. 34).

In a later work by one of the authors and her further collaborator, a proposal for combining activity theory and semiotic theory has been made (Bödker and Andersen 2005). In the referred paper the authors make an important effort to put the mediated action model of CHAT and a semiotic interpretation of that model in connection to each other (without direct reference to Peirce). The authors concentrate on showing that via the semiotic interpretation of action it is possible to understand the communicative function of instruments, and they demonstrate the use of this point for interface design. The arguments of the authors correspond very much to what we have argued on the three functions of tools. Bödker and Andersen do, however, concentrate on verbal communication as actual form of interpretation, but do not consider habit as a possibility to understand meaning in the functioning of a system, the focal topic for understanding joint system functioning.

3.4 Articulating the similarities and differences of the three approaches

In this section we shall make a summary of the above three approaches that share the intention of redefining the unit of analysis in the study of human conduct in the connections of CSE. This summary is provided in a form of a table (Table 2) for clarity reasons. The table deals with the main question of this chapter i.e. what the joint system is, and how its elements and boundaries are defined in each of the three variants discussed.

Table 2 Comparison of three joint system approaches of cognitive systems engineering
Full size table

Table 2 indicates that the Risö/CHMI approach and ours share the methodological bases. Both have a close connection to the CSE tradition, but they apply praxis-oriented methodologies like pragmatism, phenomenology, and Marxist praxis concept (that lies behind the CHAT) to complete the CSE. Hollnagel and Woods who were central figures in the development of CSE, provide an extensive description of the emergence of CSE and their own thoughts in their above referred two volumes. The authors make reference to some of the praxis-oriented methodologies, but a more analytic line of thought, systems thinking, is finally preferred.

Having a connection to CHAT, ecological psychology and their methodological sources the Risö/CHMI and our approaches understand the unit of analysis of joint systems to be a human–environment system that is mediated by technology, i.e. activity. Because the environment provides the object and purpose for activity, the Risö/CHMI and our approaches are focused on concrete features of the environment, and lean to a context-dependent analysis of joint systems. In Hollnagel and Woods case the general system theoretical frame does not insist on emphasising intentionality of action. Hence context-independent approach is preferred. This is connected with the choice of preferring human–technology system as the unit of analysis.

Adaptation is one of the central interests in the JCS approach, and the theory focuses on defining patterns of joint system functioning. We share this interest with Hollnagel and Woods. Search for a vocabulary to describe generic tendencies in a joint system is not expressed in the Risö/CHMI approach. For Hollnagel and Woods patterns are kind of system demands for adaptive behaviour. The types of demands are identified, and they are resilience, coordination, and affordance. Because these functional patterns are identified with system theoretic generic terms, new vocabulary concerning human conduct is not actually used, nor needed. It is also evident that the concept of meaning is not necessary to define patterns. These are probably the reasons why, notwithstanding the clearly expressed aim to analyse how the joint system functions, the authors appear to be attached to the concept of action when describing the joint system internal functioning. All in all the joint system’s way of functioning is viewed externally, as a demand that designers have to meet when creating such systems. In our case, meaning is central in defining patterns because we comprehend patterns as joint systems’ internal regularities that explain their behaviour. The concept of habit is a tool for us to identify meaningful patterns.

As indicated above, Hollnagel and Woods are interested in the design of joint systems. So are we. The Risö/CHMI approach design interest is focused on environmental characteristics or affordance spaces. Patterns are more or less direct objects of design for Hollnagel and Woods. We find the situation more complicated. We see that the joint systems emerge and change through their functioning, and during this process adaptive internal patterns are formed. We see, however that it is possible to talk about designing of joint systems. In order to be capable of developing adaptive functioning, the object of design should be to create prerequisites like those that Hollnagel and Woods consider as patterns of CSE. We prefer to use the expression of designing adaptive practices of joint systems.

4 Joint intelligent system design process

In this chapter we shall return to our starting point, the design activity itself. The question we put ourselves is how the joint intelligent system could be produced during a design process. We have indicated that a vision of a ubiquitous intelligent environment creates new research problems that the joint intelligent system design approach should tackle. While “immediate design” requires consideration of usage processes, “remote design” needs new theories and concepts to improve understanding of usage and human action from a joint system perspective. These two modes of design complete the “product design” mode which has not lost relevance, since products are the material results of the design process and respond to the immediate needs and wishes of users. Remote design provides new possibilities and prerequisites for design. Also, fulfilling the knowledge demands of all three design modes requires acknowledging the validity of both Modes 1 and 2 types of knowledge production.

The basis of our JIS design approach is built on the two central ideas mentioned above: the modes of design that were identified in our research project on the design of intelligent environments (Kaasinen and Norros 2007; Keinonen 2007), and the two modes of knowledge production that were characterised in chapter two by reference to Gibbons et al. (1994). The JIS design process model that is depicted in Fig. 3 has been influenced also by the model of Hollnagel and Woods (2005, p. 5) named “practice-centred approach to research” which in a pertinent way summarises the knowledge production process in joint cognitive system design. In this connection we would also call attention to another design process model that has been of interest for us when framing the JIS design approach. The method proposed by Elm et al. (2003) called “applied cognitive work analysis” is consistent with the Woods and Hollnagel model but takes a much more practical approach to design.

Fig. 3
figure 3

The joint intelligent system design process

Full size image

We comprehend the joint intelligent system design model to be structured by two dimensions that interact in the design process. The first dimension can be described by its two poles that symbolise the modes of knowledge creation. The other pole indicates the urge to generic regularities and well-formulated knowledge, which is the traditional aim of scientific knowledge production. This rationale for knowledge production corresponds to Gibbons’s Mode 1 knowledge production. The opposite pole denotes knowledge production that emphasises contextual and practical knowledge and implicit insight, so-called tacit knowledge. This type of knowledge production was described by Gibbons as the emerging Mode 2 knowledge production.

The second dimension is stretched towards two poles that may be described by the new modes of design that supplement traditional product design: remote design and immediate design. The immediate design focuses on local and immediate user needs and experiences. It aims at the development of products that could fulfil the needs and at providing possibilities that users can utilise and shape into practices. Hence, it heavily relies on user participation in design. The remote design is more distant in the sense that it aims at abstracting from the immediate and, instead, creating general solutions that offer possibilities and prerequisites for the future. Remote design involves people who are concept and science oriented and interested in creating formative solutions.

The two dimensions frame a design space that is divided to four fields each representing a different design activity. The nature of the design activities may be defined by both the dimensions and the poles that they are closest to. The design process that proceeds by moving between the design fields. Since the design activities provide qualitatively different viewpoints to design we maintain that the design process is progressive rather than iterative.

The outputs of the design activities are twofold: the first outcome type is the results of immediate design that consist of local solutions such as new products and practises; the second outcome type is the results of remote design, i.e. enabling solutions and new possibilities or “design seeds” as aptly labelled by Woods and Hollnagel (2006). The two different types of outcomes may be supported theoretically by the ideas of the cultural-historical theory of activity and Järvilehto’s human–environment theory (Chap. 3) according to which the outcomes of the joint system (that design also is) are, on the one hand, material products and, on the other hand, possibilities to develop the current activity or create new ones.

Design artefacts are a concrete way of explicating the knowledge created in each design phase and conveying the results to the next phases. The notion of “design artefact” was borrowed from Elm et al. (2003). As Elm et al. say “design artefacts capture the results of each of the intermediate stages in the design process. […] They form a continuous design thread that provides a principled, traceable link from cognitive analysis to design” (Elm et al. 2003, p. 362). It is evident that in the practical and product-oriented design process that Elm et al. describe, design artefacts represent the generative knowledge production and the aims of remote design.

In the following we shall briefly outline the contents of each of the design fields with regard to two issues: what is the rationale of design activities in this field, and what types of methods are used. The design fields are labelled by their specific objects and tools of design. Also, the phases of the design process that map onto the design fields are described.

4.1 Practical problems

Practical problems is the design field that may often but not necessarily be the starting point of design. The rational of design in this field should be to define the zone of proximal development of the practice. We use this Vygotskyn expression (Vygotsky 1978) to indicate that problem definition should not reduce to the most apparent problems and complaints of the users. With the participation of researchers and designers the users are urged to define the problems but also to define the first step forward.

There is an abundance of methods within the user-centred design that could be used in this design field. In many cases they, however, focus on usability or usage in much too narrow terms. When the work or other activity processes are complex and technically advanced, detailed field studies of the present work situation are needed to identify developmental needs. A description of different approaches within this broader category is provided in a recent book edited by Hollnagel (2003b). Our Core-Task Analysis approach is one example of such methods (Norros 2004). Ethnographically oriented work place studies are also possible starting points for design (Bannon 2000).

In the later phases of maturation of the design process, the “practical problems” design field includes also creation of scenarios, and user evaluations of design solutions. Scenarios illustrate how the designed joint intelligent system functions in its usage context, and form a basis for requirements specification conducted in the next design phase. The evaluations typically support design and are incorporated directly with the particular design solution. Nevertheless, a proper and informative evaluation of joint cognitive systems is a complicated issue. This has been discussed widely e.g. within the nuclear power plant field in the context of control room integrated validation (see a review in Norros and Savioja 2004). For this purpose we have developed a methodology that corresponds to the ideas expressed in this paper (Savioja and Norros 2008).

4.2 Design problems and joint system requirements

The first task in this design field is to formulate practical problems into design problems that will be tackled in the rest of the design process. The rationale of design activities in this field is characterised by the urge to generalise from the field studies and user experiences. In this field the generalisation does, however, not mean escaping from the domain context in which the designed system is embedded. On the contrary contextual domain modelling should be accomplished in order to define adequately the joint system requirements.

The “Applied cognitive work analysis” method by Elm et al. (2003) is an excellent example of requirement analysis. The authors utilise the concept “functional abstraction network” (FAN) and acknowledge an earlier work of Woods and Hollnagel (1987) as one important source of this method. In modelling tasks we find important that not only the intrinsic demands of the domain but also the psychological or cognitive user demands are considered. In our core-task analysis approach we have developed a procedure to map the generic work domain demands and the demands relating to acting, knowing and collaborating that we defined as generic psychological demands of adaptive behaviour. As a result we acquire core-task demands which are used in the further phases of analysis and design. The core-task demands and the demands for adaptation or patterns as described by Woods and Hollnagel (2006, p. 8), namely coordinated activity, resilience, and affordance, seem to have a close resemblance.

We have found the concept of pattern useful and adopted the term “adaptive patterns” to describe the adaptive ways of functioning of a joint system. On the one hand, adaptive patterns are systems’ inherent ways of acting that are meaningful and thus worth repeating, and that continue developing during the life-time of the system. On the other hand, adaptive patterns are a means for designers to identify, generalise, and describe the “intelligence characteristics” i.e. the principles according to which the adaptive functioning of the joint system is achieved. In the early phases of the design process adaptive patterns function as design artefacts as they can be used to convey knowledge of the ways of functioning to the next design field. Another type of design artefacts of this design field that is used later in the design process is specifications of joint system requirements.

4.3 Concepts and theoretical explanations

The rational of this design field is methodical. We do not consider generic models as the ultimate purpose of research and design but rather see that generalisations are tools for explaining new actual phenomena and tackling design problems. Methods incorporate experience and interpretations that aid in approaching interesting new phenomena. Typical methods, design tools, and outputs of this design field are e.g. design patterns, platforms, standards, and algorithms.

In this design field it should be important to reflect on the design process and its rational and to consider the adequacy of the methods. As a matter of fact, the present paper and issues it discusses could be considered an example of activity that belongs to this design field. As part of the reflection on methods, comparisons over domains are important as they offer the possibility to discuss the validity and range of the methods, models and concepts.

Because this field is also bounded by the design dimension and it touches the “remote design” pole the rational of a method-orientation should be also interpreted in design terms. From this perspective product conceptualisation and platform design issues are to be considered as activity of this design segment.

4.4 Experimenting and solutions

The rational of this design field is the development of solutions and experimentation with them. When creating new solutions, designers’ know-how of technical and methodical possibilities is utilised in making sketches of possible design solutions. On the basis of both the knowledge of present practices and the new ideas generated in the earlier design phases, it is possible to create a concept of operations that articulates the vision of the new product or activity by using scenarios and other visualisations. The concept of operations, the notion of which was borrowed from (Fairley and Thayer 1997), functions as a design artefact that conveys the results to the next design field.

In a later phase of the design process, experimenting may take place by simulations or tests with actual technical prototypes. Independent of the nature of prototyping an important issue is how the results are gathered, analysed and, in particular, evaluated. It may be proposed that in this design field, in addition to the prototypes and solutions, evaluation metrics could be central design artefacts. The criteria of evaluation emerge from the work accomplished in all previous phases. In this phase decisions have to be made concerning the metrics and the precise evaluation criteria that apply to the joint intelligent system prototype.

5 Concluding remarks

In the beginning of this paper three design rationales for human factors were outlined. The third of them was characterised by being aimed at creating new possibilities and uses of living. This idea was borrowed from Hancock and Chignell (1995), two authors who advocate an ecological approach to design, and are proponents of joint systems approach. We made an attempt to find arguments to support the notion of a joint systems approach. By providing evidence of trends in design we demonstrated that new problems for new theoretical approaches are emerging, and a paradigm shift becomes more and more evident in CSE. This conclusion is also supported by the fact that different versions of joint system approaches are created by researchers who work more or less independently of each others. In the present paper we discussed three such approaches. These approaches have, by no means, developed independently, since a common background in CSE and many concrete ties over the years connect them. Two of these approaches demonstrate an interest in identifying generic regularities or patterns that would describe how a joint cognitive system functions. A scrutiny of the principles according to which a joint system is formed revealed that notwithstanding similarities between the approaches, regularities and adaptive principles are sought from two distinct perspectives. One is to work on systems theoretical vocabulary and identify structural patterns in the system. The other way is to consider the system as human activity, take an actors’ point of view and reveal cultural regularities within people. We prefer the latter strategy. When following this solution it becomes necessary to develop empirical concepts with which to approach joint cognitive systems. The semiotic concept of habit is the central means there. Habits were shown to be regularities or patterns of behaviour that represent joint cognitive system functioning. Habits elaborate the communicative function of technology and connect action to its meaning via reflective and non-reflective embodied processes. It appears that semiotic or pragmatic approaches combined with other approaches will become a serious option in the CSE in the future.