Journal Information

Article Information

A framework for knowledge innovation


The study of various knowledge creation approaches (like operations research, systems thinking, cybernetics, complexity, knowledge management, and scientific method) leads to the development of a framework describing a sufficient capability for trans-disciplinary knowledge innovation and knowledge creation. This is done by looking at knowledge creation as a complex phenomenon, and using an appropriate approach, as found in the Cynefin Framework. The proposed knowledge innovation framework highlights the various aspects to be developed in order to define a capability which can be used for knowledge innovation, scientific problem solving, and quality assurance for knowledge work. This framework can be seen as a ‘knowledge technology’ that can be developed and implemented like any other technology.

Knowledge creation/discovery

This paper does not wish to enter the debate about knowledge creation vs. knowledge discovery. The term knowledge creation is therefore used for any of these two, denoting the uncovering or creating of new knowledge, with new knowledge simply being knowledge not previously known.

Even a superficial look at the results of scientific work through the centuries presents its unparalleled success in the pursuit of gaining understanding and creating new knowledge. This historical success story did not inhibit the strong challenge against the ‘restricting’ scientific method in the twentieth century. This challenge came about precisely because the proponents of science tried to extrapolate the success of scientific method beyond the limits of the natural world into social systems like management and organizational sciences. In these spheres the ‘classical Newtonian scientific method’ proved to be less successful, and some costly failures resulted.

As a result new approaches developed to gain understanding of management and organizational realities and to tackle the problems presenting themselves in those fields (e.g., operations research, systems thinking, cybernetics, knowledge management, complexity theory). While these approaches do relieve the pain of organizational and management decision makers, the question remains whether or not they will be seen in the future as having the same, or even a higher, level of success in creating new knowledge in their spheres of application than the scientific method has proven itself to have in the natural world.

This question can be translated as: what are the salient features of scientific method that facilitated its success in knowledge creation in the natural world? If these salient features can be identified and applied appropriately to the approaches used in management and organizational knowledge creation, then the probability of repeating the historical success of scientific method might be increased. Is this possible?

The first obstacle is in finding what those salient features are. A study of the different views of scientific method over the centuries does not provide a uniform description which can be used as basis. In fact, the century-long debate about the method and description of ‘scientific method’ stands in stark contrast with the success of science itself—it does not provide a description of such a set of features. This could lead one to abort the project of finding them straight away, giving up the hope of repeating the success story in other fields. One way of trying to overcome this deadlock is to recognize that scientific method itself is a complex system, and therefore needs to be studied as such.

The next obstacle is the fact that there is no agreement on the nature of complexity. The two dominant schools of thought regarding complexity are deterministic complexity and non-deterministic complexity. These two are related, but differ sufficiently in their basic point of departure that they provide alternative ways to approach the problem at hand. Noting that deterministic complexity is mostly adhered to by the natural sciences and non-deterministic complexity by social sciences, and recognizing that the problem at hand is inherently a social system, and that the results will be applied to fields in social science, the decision is made to approach the study to identify the salient features of ‘scientific method’ as a non-deterministic complex system, without entering into the debate of deterministic vs. non-deterministic complexity.

Non-deterministic complex systems

The next obstacle is to understand what a non-deterministic complex system is, and to confirm the decision to approach the study as an non-deterministic complex system.

The work of Cilliers (1998) provides a definition of non-deterministic complex systems without referring to deterministic complexity at all. His list of attributes for a complex system is (Cilliers, 1998: 119-123):

  1. Complex systems consist of a large number of elements;

  2. The elements in a complex system interact dynamically;

  3. The level of interaction is fairly rich. By this is understood that not all interactions are of the same nature, and that the level of interaction are growing or diminishing dynamically;

  4. Interactions are nonlinear. Cilliers holds that nonlinearity is a precondition for complexity. He states that “linear, symmetrical relationships give rise to simple systems with transparent structures” (Cilliers, 1998, 120). He describes the social system as inherently nonlinear;

  5. Interactions have a fairly short range. Interactions are usually with other elements around them, although that should not necessarily be understood in geographical terms. Since there is no meta-level controlling the interrelationships, the behavior of a complex system is best described in terms of a multiplicity of local ‘discourses’ (Cilliers, 1998: 121). He distinguishes between short range interactions and long range influences, stating that local interactions can have long range influence;

  6. There are loops in the interactions. Feedback is an essential aspect of complex systems. These can be intricately interlinked loops, also feeding back to themselves;

  7. Complex systems are open systems;

  8. Complex systems operate far from equilibrium. A constant flow of energy is necessary in order to change, evolve and survive as complex entities. As such complex systems survive as a process, being defined by what it is doing, rather than by its origin or its goals;

  9. Complex systems have histories. The history of a complex system is not an objectively given state—it is a path distributed over the system and is always open to multiple interpretations;

  10. Individual members are ignorant of the behavior of the whole system in which they are embedded. “Single elements cannot contain the complexity of the whole system and can therefore not control or comprehend it fully” (Cilliers, 1998: 122).

These attributes provide an explanation of why the description of scientific method remains illusive even after centuries of study. None of the attributes listed above contradict any of the attributes noted about scientific method through the centuries, while together they give reason to believe that the hope of achieving any one description of the scientific endeavor might be a futile hope indeed. Complexity, however, needs to yield more value than to provide a legitimate disclaimer to be of use for practical problem solving, an issue which this paper hopes to illustrate.

Cynefin framework

The next obstacle is finding an appropriate technique in non-deterministic complexity to study scientific method in order to identify its salient features. Literature offers a number of techniques, but the Cynefin Framework (Kurtz & Snowden, 2003) is used here as basis, since it provides specifically for the understanding of the attributes of complex systems in its non-deterministic definition in a way that complements Cilliers’s view, while it also gives a way to study such problems—through ‘perspective filters’, managing the emerging patterns. When they describe the domain of complexity, making human behavior prominent, they say:

“This is the domain of complexity theory, which studies how patterns emerge through the interaction of many agents. There are cause and effect relationships between the agents, but both the number of agents and the number of relationships defy categorization or analytic techniques. Emergent patterns can be perceived but not predicted; we call this phenomenon retrospective coherence. In this space, structured methods that seize upon such retrospectively coherent patterns and codify them into procedures will only confront new and different patterns for which they are ill prepared. Once a pattern has stabilized, its path appears logical, but it is only one of many that could have stabilized, each of which would have also appeared logical in retrospect. Patterns may indeed repeat for a time in this space, but we cannot be sure that they will continue to repeat, because the underlying sources of the patterns are not open to inspection (and observation of the system may itself disrupt the patterns). Thus relying on expert opinions based on historically stable patterns of meaning will insufficiently prepare us to recognize and act upon unexpected patterns” (Kurtz & Snowden, 2003).

It is important to note the warning that using the retrospective coherence of emerging patterns to make sense of future events, may prove restrictive in recognizing unexpected patterns. This limitation is recognized in this study as an inherent feature of complex systems, and is therefore accepted as an inherent weakness of the proposed set of salient features of science to be extracted from the study—they represent a minimum sufficient set of features, not a complete set.

At the same time the self-organizing attributes of complex systems, with emergent properties giving expression to underlying sources of the patterns not open to inspection (according to Snowden), provide the impetus for using perspective filters to study complex systems. Looking at a complex system from a sufficient set of perspectives provides the best way to understand the system, even though this understanding is partial and provides a snapshot of the system in time. Three other warnings about the results of the study can be articulated:

  1. The results presented here do not provide a conclusive answer to the problem. They are very dependent on the choice of perspectives to study, providing a partial solution, yet striving to provide a sufficient solution by choosing a sufficient set of perspectives;

  2. The results presented here need to be taken into reconsideration from time to time as the system is changing dynamically and new or other patterns might develop;

  3. Observing the retrospective patterns are in essence an interpretation and multiple interpretations can be developed depending on the a priori knowledge and assumptions. (One example is rewriting history textbooks when a major political power shifts takes place politically). The obvious assumption in this study is the observed unparalleled success of scientific method in knowledge creation in the past.

All of these warnings are common to all complex systems, not specific to the problem at hand. They are given to point out the need for ongoing study to confirm, refine or change the results proposed here.

Choosing the perspectives to develop

The aim is to study knowledge creation approaches in comparison with scientific method in order to identify the salient features of scientific method that facilitate its productiveness in knowledge creation. For this reason the following criteria are used for inclusion of an approach as a perspective to study:

  1. The field claims to be of value for practical problem solving;

  2. The field claims to add value to the process of knowledge creation.

The inclusion of approaches claiming to add value in both problem solving and knowledge creation is done because it is recognized that knowledge is often created by solving a problem (Popper, 1963). They are:

  1. Scientific method;

  2. Operations research;

  3. Systems thinking;

  4. Cybernetics;

  5. Complexity;

  6. Knowledge management.

Developing the perspectives

A detailed discussion of the development of each of these, with specific focus on the latter part of the twentieth century, is given elsewhere (Van der Walt, 2005). Only a summary of each is given here.

1. Scientific Method: Positivism proved to be the dominant view of the twentieth century, even though its dominance has been severely challenged by Popper. Monism was the only view of scientific method assumed through the centuries, but in the latter part of the twentieth century pluralism became dominant. Pluralism presented itself at first as relativistic and behavioristic (e.g., the views of Kuhn and Laudan). Feyerabend was probably the best example of how deeply entrenched monism was—when he (reluctantly) came to the conclusion that there was no one scientific method, he concluded that there was none (Feyerabend, et al., 2001).

Towards the end of the twentieth century Popper received more favorable readings while Kuhn’s views were increasingly challenged (e.g., Nola, et al., 2001). While Popper was a monist, his articulation of scientific method provides for application in both the natural and social sciences. Popper’s view of scientific objectivity, especially, proved to be instrumental in unlocking value during the study of the perspectives. A summary of his view is provided:

Popper on objectivity

Popper’s view of objectivity is that our theories can be articulated in language, and as such can be understood and criticized by others.

His view that objective, rational knowledge is inherently fallible—and that we can never justify, but only criticise, it—is essential to Popper’s philosophy of science.

Popper believed, like Kant, that we do have a priori theories when we go about our empirical work, but unlike Kant he did not believe that our a priori knowledge is infallible (Notturno, 2000). He believed rather that such knowledge is fallible and the aim of our empirical work is to test our a priori theories. It is important to note that Popper placed a high value on empirical testing of theories, but did not subscribe to deriving facts from empirical work, unless we can do so by deductive argument.

In this way Popper opened the way for science not to have a view of objectivity as “knowledge without context and devoid of human interpretation/opinion” but rather as an articulation of our theories in a way that would make our inherent subjectivity visible and therefore open to criticism. It is this change that makes Popper’s work of much value in the pursuit to build the bridge between natural and social sciences—a strength he himself already exploited and illustrated.

2. Operations research: Operations research was born during the Second World War when science was used by wartime decision makers to solve problems, and assist them with decision making. The success of this led to the belief that science can also be used in management and business/organizational decision making and problem solving (Callahan, 2002).

An instrumentalist expression of this endeavor was pursued in the UK, called operational research, while the early adherence to scientific method made way in the USA for a more holistic, systemic approach, called operations research. This led to the birth of systems thinking and cybernetics. A good example of this migration away from ‘scientific method’ to a holistic approach is Russell Ackoff, who became a management science guru (Ackoff, 1962, 1979).

3. Systems thinking: The limited success of scientific method in management and organizational studies led to the view that the whole system needs to be modeled, not only isolated problems. This introduced the development of systems thinking where coherent mental/conceptual representations of the whole and their interactions were modeled instead. This expansive approach proved to be unproductive, and in the latter part of the twentieth century a return to ‘appropriate’ boundaries around ‘soft systems’ is proposed, coupled with ‘appropriate reductions’ of reality to make the study of the ‘action space’ more manageable. Robert Flood is an example of the migration back to scientific discipline, as illustrated in his book Rethinking the Fifth Discipline (Flood, 1999):

  1. For objectivity he proposes ‘triangulation’ and ‘recoverability’. Together these two represent a close match for Popper’s view of objectivity, obviously assisting with overcoming the difficulty that in social systems objectivity cannot be ensured by requiring repeatability, as is the case in natural sciences experimentation;

  2. He provides a way to study and articulate ‘appropriate’ boundaries and boundary judgements (initial values);

  3. He develops four sufficient ‘windows’ (perspectives) for organizational problem solving, namely: i) Systems of structure, ii) Systems of processes, iii) Systems of meaning. and iv) Systems of knowledge power. In this way he provides an ‘appropriate’ reduction of the organizational reality to be studied.

Through doing this he illustrates the movement back to ‘appropriate’ boundaries, ‘appropriate’ reduction, ‘appropriate’ initial values, and an insistence on articulating the process we follow sufficiently that our decisions (subjectivity) can be criticized, which is scientific objectivity according to Popper.

4. Cybernetics: Two views of cybernetics became prominent, namely the input—processing—output—feedback model (from mechanics) and the autopoietic model (from biological systems). Heylighen, et al. (1990) proposed that in organizational realities the main difference between these two is where the observer is situated. The mechanical view is visible to an external observer and the autopoietic view for an internal observer. It was therefore not possible to escape from having to position the observer in order to gain understanding of the problem, so that initial conditions and assumptions are inevitable.

Cybernetics and general systems theory interacted and pursued an articulation of reality from their respective views, maintaining a holistic view instead of the ‘reductionist’ view of scientific method. Although they succeeded in creating new insights, the problem of positioning of the observer became increasingly problematic. In this way the need for ‘appropriate’ boundaries and reductions of the problem space presented itself once again to facilitate meaningful study.

5. Complexity: Complex systems were first defined by Bohr and his team in physics. This description of complexity is deterministic in nature, but so complex and involving so many different nonlinear interactions that it is not possible to model the system faster than it actually developing in real life. Studying weather patterns is a common application of this view.

Social science borrowed the definition of complex systems and changed that definition at the same time to be non-deterministic. In this articulation the main attributes of complex systems as described in deterministic complexity are kept, except that the dependency on very accurate initial values is discarded, claiming that the non-deterministic nature of the systems make them insensitive to the initial values.

In non-deterministic complexity a system can only be described by a model at least as complex as the system itself, so that modeling a complex system becomes impossible. Any partial model will always yields incorrect results in unforeseen ways, due to the nonlinear nature of the system. While this adds much value to understanding complex systems and complex problems, it needed to move beyond being a disclaimer. To do that ‘appropriate’ boundaries and ‘appropriate’ reductions need to be found to model the system partially, knowing and accepting the inherent weakness of such modeling. This study is one example of such a model, with the perspectives proposed as an articulation of such an ‘appropriate’ boundary.

The main value added by studying a complex system is its self-organizing nature and emergent properties (patterns), while the system itself strives to remain in a state of homeostasis. A complex system is therefore more robust than non-complex systems. Studying these in a meaningful way presents its own difficulties, not the least being the challenge that observing emergent properties are influenced by a priori preferences, assumptions or knowledge. This is accepted in this study, and once again highlights the requirement of articulating any such preferences or knowledge to a level where they can be criticized if the study is to be objective.

6. Knowledge management: Knowledge management was initially seen as ‘getting the right information to the right person at the right time’. It was soon recognized, through the work of Polanyi and others that tacit knowledge was not taken into account in that view, because knowledge was seen as completely inanimate.

The next phase of knowledge management then became to develop techniques which would enable and stimulate people to articulate their tacit knowledge. Techniques like mind maps fall into this category. Knowledge is still seen as the articulated, captured knowledge that can exist independent of the agent producing it in this phase.

It was then recognized that articulating tacit knowledge was a very costly (time-consuming) effort which involved highly skilled people (and therefore also costly in monetary terms), while the results are only useful for a limited period of time due to the changes in the context. In this way stagnation takes place if knowledge articulated in this way is used ‘beyond its shelf life’, or a costly exercise needs to be repeated.

The third phase of knowledge management started as a result, with the emphasis moving away from articulating knowledge to rather utilize the tacit knowledge by involving the agent having the knowledge directly in the process (McElroy, 2002). In this way managing knowledge becomes managing the knowledge processes rather than focussing on articulating the knowledge, or the management of the articulated knowledge. The quest then becomes to create and manage the processes that would facilitate knowledge creation and utilization. This study claims that it is in this phase that knowledge management can benefit from the insights of scientific method, because knowledge creation (and the process for doing so) was the domain of scientific endeavor for centuries already.

Identifying the salient features of knowledge creation

A detailed discussion of how these features were identified is given elsewhere (Van der Walt, 2005). A short summary of the argument is given here together with a list representing a minimum list of features for productive knowledge creation.

The assumption was made that it was the restrictive ‘reductionist’ nature of scientific method that caused ‘scientific method’ to have limited success when applied to fields other than the natural world, so a holistic approach was pursued. In time the expansive nature of the resultant holistic approaches had to be reduced to be more productive if real problems were to be solved. The way in which the various approaches coped with this awareness provides a confirmation of some of the features commonly known in scientific method. They can therefore be proposed as important candidates for inclusion in any approach wanting to be productive in knowledge creation. They are:

  1. Objectivity;

  2. Boundaries;

  3. Initial values (including boundary judgements and placing the observer);

  4. Assumptions;

  5. Reductions.

The one word that makes them applicable to all the knowledge approaches, and not only those successful in the natural world, is the word appropriate. It can now be seen that these were appropriately defined for the natural sciences, and what was appropriate for certain aspects the natural world (which is a complex system itself), was not appropriate for other contexts. Identifying appropriateness has now been identified to be of primary importance when pursuing knowledge creation.

The next obstacle is how to find appropriateness in any given problem case to be explored.

Problem types and appropriateness

This study has already alluded to the proposed definition of how to search for appropriateness, namely to look at the type of problem to be studied. In this paper scientific method was defined as an non-deterministic complex system and studied by a technique developed specifically to study problems having that nature.

Literature already proposes such problem categorization frameworks, of which the Cynefin Framework (Kurtz & Snowden, 2003) is one. Another one is the system of systems framework proposed by Flood and Jackson (1991). The necessity of appropriateness is therefore already recognized in practical problem solving.

Apart from the empirical observation that the features of scientific method were not appropriately applied in fields other than the natural world when scientific method was extrapolated to those areas, formal indicators also point in the direction of pluralism:

  1. Development of different proof theories in logic during the twentieth century pointed to the importance of different problem types and that they should be approached differently;

  2. Gödel’s Second Incompleteness Theorem shows that (at least in Hilbert spaces) a system cannot be described coherently and completely at the same time.

The proposed model for disciplined knowledge creation

Figure 2 provides a conceptual model of the proposed model. Both levels of abstraction, namely the Problem/Context Categorization Framework (PSF) and the Scientific Problem Solving Process (SPS) need to be appropriate for the context they are to be used for.

The PSF needs to be coherent and able to categorize all problem types which present themselves from the context, while the SPS is a complex process with multiple, dynamic feedback loops. Its specific application for any given problem stems from the problem type, as well as the values and natures of all the features for scientific knowledge creation defined above.

Case study: Institute for Maritime Technology

The case study illustrates the usefulness of this model for research management. It can also be used for:

  1. Scientific problem solving and knowledge creation;

  2. Quality assurance for knowledge work.

The Institute for Maritime Technology (IMT) is an R&D institute servicing the South African Navy (SA Navy) with a scientific support service. Part of its mandate is to provide decision support to assist the SA Navy with fact-based decision support. The definition of what capabilities to develop for providing decision support is problematic due to the following:

  1. From experience it is known that it takes about three years to develop a capability from nothing to a point where practical utility can be provided to the SA Navy;

  2. Due to constraints in resources it is always the reality that the decision support group is confronted with more problems than it knows how to solve. Being able to provide an acceptable service within those constraints requires careful planning and development of capability;

  3. The necessary multi-disciplinary nature of the required decision support service complicates the interpretation of the mandate to provide a scientific service;

  4. The fact that decision making is a social activity makes the interaction and integration with the rest of IMT (which focuses on natural science and technology development) a constant concern for the decision support group.

By using the model developed above all of these difficulties were overcome, with a robust definition of what capability to develop, how to ensure high standards of scientific work across scientific disciplines, and how to populate the processes of IMT in a way that adds value for knowledge work—even though they are specifically designed for technology (hardware) development and not for knowledge work.

The most pressing requirement, namely not to waste time developing capabilities that become unimportant to the client (due to changes in the context) even before they are able to provide utility, is overcome by focussing the attention on defining appropriate problem types, developing the decision support practitioners in appropriate diagnostic and synthesis skills, and building a professional network with the appropriate analytical (modeling) expertise with only a sub-set of most common areas of modeling expertise developed and maintained inhouse.


Studying the philosophy of science as a complex system in comparison with the development of other contemporary knowledge creation approaches proved to be very productive and yielded results that would not have presented themselves if any of these were studied individually.

These results made it possible to derive a realistic pluralist model for scientific problem solving, utilizing the categorization of problem types to approach the variety of problems which could present in a productive way. This model has already proven to be of practical value in a trans-disciplinary case study where scientific problem solving has a multi-disciplinary nature.



Ackoff, R. L. (1962). Scientific method: Optimizing applied research decisions, New York, NY: John Wiley & Sons, ISBN 0898746612 (1984).


Ackoff, R. L. (1979). “The future of operational research is past," Journal of the Operational Research Society, ISSN 0160-5682, 30(2): 94-104.


Cilliers, P. (1998). Complexity and postmodernism: Understanding complex systems, London: England, Routledge, ISBN 0415152879.


Callahan, D. (2002). Kindred spirits: Harvard Business School’s extraordinary class of1949 and how they transformed American business, New York, NY: John Wiley & Sons, ISBN 0471418196.


Feyerabend, P., Lakatos, I. and Motterlini, M. (eds.) (2001). For and against method: Including Lakatos’s lectures on scientific method and the Lakatos - Feyerabend correspondence, Chicago, IL: University of Chicago Press, ISBN 0226467759.


Flood, R. L. (1999). Rethinking the fifth discipline: Learning within the unknowable, London, England: Routledge, ISBN 0415185300 (2002).


Flood, R. L. and Jackson, M. C. (1991). Creative problem solving: Total systems intervention, Chichester, England: John Wiley & Sons, ISBN 0471930520 (2002).


Heylighen, F., Rosseel, E. and Demeyere, F. (1990). Self-steering and cognition in complex systems: Towards a new cybernetics, Abingdon, England: Taylor & Francis, ISBN 2881247296.


Kurtz, C. F. and Snowden, D. J. (2003). “The new dynamics of strategy: Sense-making in a complex and complicated world," IBM Systems Journal, ISSN 0018-8670, 42(3): 462-483.


McElroy, M. W. (2002). The new knowledge management: Complexity, learning, and sustainable innovation, Burlington, MA: Butterworth-Heineman, ISBN 0750676086.


Nola, R. and Sankey, H. (ed.) (2001). After Popper, Kuhn and Feyerabend, New York, NY: Kluwer Academic Publishers, ISBN 1402002467.


Notturno, M. A. (2000). Science and the open society: The future of Karl Popper’s philosophy, Budapest, Hungary, Central European University Press, ISBN 9639116696.


Popper, K. R. (1963). Conjectures and refutations: The growth of scientific knowledge, 5 edition, London, England: Routlege, ISBN 0415285941 (2002).


Van der Walt, J. M. (2005). “Technology for knowledge innovation: A realistic pluralist scientific problem solving capability,” University of Pretoria,

Article Information (continued)

This display is generated from NISO JATS XML with jats-html.xsl. The XSLT engine is Microsoft.