Chapter 5: Organizational Learning in Machine Bureaucracies
5.4 Consequences of Machine Bureaucracy for Organizational Learning
5.4.3 Machine Bureaucratic Single-Loop and Double-loop Learning
organizational learning norms. It takes its results (learning norms) as an internal contingency factor for analyzing the effectiveness of the single-loop and double-loop learning processes and the role and value of MICS. The reason for this limition is that MICS has no role in the deutero learning process, because MICS is part of the (procedural) learning norms itself.
This subsection explores the role of machine bureaucracies in the development and removal of management theories, their storage, use, dissemination and adaptation. These learning activities are labelled double-loop and single-loop learning.
1. Double-loop learning processes Development of management theories
Stable and simple environments do not require much double-loop learning. The development of knowledge is then often delegated to specialists in the technostructure (frequently organized in a Research and Development department), sometimes with the involvement of the management as well. For instance, in an insurance company I visited21, the principle trend for managers was to delegate all
knowledge development processes to specialists, even the most simple ones. Development of knowledge is done infrequently and has a low priority in simple and stable environments. Feedback comes slowly and irregularly, mostly organized in market research. In these studies potential customers express their opinion of products and services and their future demands when requested. The results of these studies lead to decisions about developing,
manufacturing and selling new products. Because of the low complexity and dynamics the results are very precise, so that good calculations of costs and benefits can be made. The environment is low risk, and therefore demotivates the search for innovations.
When the environment becomes more dynamic and complex a more active knowledge development approach is required. Besides, it becomes difficult for specialists alone. Often in machine bureaucracies problems are encountered in the manufacturing of new products because members from the production department were not involved in the product design (Hill, 1984). The risk also increases as the environment becomes more complex. Processes of critical evaluation must be speeded in order to gain relevant knowledge, because knowledge depreciation also speeds up. Knowledge development must also become systematic (e.g. explicitly asking clients about satisfaction, and actively searching for problems to be solved). When the dynamics of the environment is very high, knowledge depreciation is faster than the knowledge development process. In this case the value of knowledge declines rapidly, and management is either left over to good luck or beome able to avoid uncertainty by creating a negotiated environment (Cyert and March, 1963). Figure 5.3 shows that the increase of the value of knowledge is directly related with the complexity of the environment. Figure 5.4 demonstrates that the accumulation of knowledge is effective until the environment passes a certain level of dynamics. After that moment the depreciation of knowledge goes faster than its accumulation of value. Speeding up the feedback process can increase the value of knowledge, as is shown in fig. 5.5.
In case of high complexity and dynamics, a delegating style is appropriate, because very short communication lines are required and much knowledge is decentralized. The decentralization also leads to a restriction of the area that must be understood, and thus simplifies the problem. This can of course lead to suboptimization and dysfunctional
effects in the longer run. It is typical of double-loop learning that it detects these suboptimization problems and solves them by generating an awareness of limitations, and the new insights that are required. Action norms (motivation to rethink the management theories especially in a broader perspective) and procedural norms (creating communications and activities to detect limitations and discover a wider perspective) must be set so that double-loop learning activities emerge.
Removal of management theories
The conservatism and reification resulting from the stability and simplicity of the organization's environment, makes the removal of obsolete procedures, rules, norms etc. especially difficult. Because an interchange of ideas and knowledge among departments in classic machine bureaucracies happens infrequently, at
best only a few people can clearly assess the impacts of a knowledge shift for the whole organization. Resistance is likely to arise from political (position power), interpersonal and socio-technical (way of working) perspectives. Slow evaluation cycles do not motivate the removal of old ways of working, because people do not see what is wrong with the existing habits (everything is fine, until real problems crop up and the process of decline can no longer be reversed). What is perceived is the risk of losing things that are valued highly, and the fact that a major shift always bears the risk of failure. This is called 'reorganization risk' in the literature (Hannan and Freeman, 1977 and 1984). Perceived low risk could easily lead to neglecting the importance of action and up-to-date knowledge. The telling style could support the change and removal of old knowledge in an autocratic way (brute force strategies). The result of problems with unlearning is that a growing tension arises between knowledge needs and the available knowledge. Management then frequently becomes mismanagement, doing precisely the wrong things. What at first seems to be an improvement, later turns out to be a failure (dysfunctional effects). As a remedy to this way of thinking, Senge therefore proposes system dynamics thinking, emphasizing the analysis of (unexpected) dysfunctions of behavior.
The factors mentioned in table 5.15 thus influence the likeliness of double-loop learning in lean and classic machine bureaucracies.
MB: Removal and double-loop trigger issues:
Lean Classic
Source of resistance to change
Much emphasis on details Expert power and position power
Perceived urgency for double-loop learning
High urgency, perceived as essential for survival and continuation of the organization
Low urgency perception
Risk awareness on shorter and longer term
High awareness of longer term risk Low awareness of longer term risk
Learning speed Higher learning and critical evaluation processes, reducing knowledge
High discrepancy of knowledge when dynamics increase (theory
depreciation in dynamic environments vs. practice)
Table 5.15: Lean and Classic Impact on Knowledge Removal. Definition of a Double-Loop Score
Many dimensions of double-loop learning are given in this subsection. We are concerned mainly with two double-loop learning activities: theory development and unlearning. In chapter 4 we found four basic fields of learning, namely: human resources, transformation (production processes), markets (for acquisition of resources and growth), products (as concrete field of productivity and efficiency). A double-loop score can now be defined as the amount of theory development and unlearning that occurs related to the four learning fields, in organizations.
2. Single-loop learning processes Storage of knowledge
In stable environments knowledge once developed in the form of procedures, norms, rules etc., becomes a person's second nature. This implies a separation between know- how and know-why. I found an example of this in a Dutch bank, where a large system of norms had been developed relating to the flow of forms and information for processing payment services. The system became so complex that many people did not know why certain forms were used and should be passed over to other departments. A consultant found out that many forms were not even applicable anymore because payment services had changed enormously in the last decades.
Sometimes only the leader or a person from the technostructure knows the connection between know-how and know-why. This knowledge inequality is consistent with the telling style of some managers, and is a source of expert power. Growing complexity requires the dissemination of know-why knowledge besides know-how, because it is difficult for the management to know everything and instruct followers effectively. Besides, tasks often require the knowledge of several people, and individual job execution is rare. In cases of high complexity, knowledge storage increases the value of knowledge because problems can be explained better and treated more effectively. The knowledge storage process however can easily lead to a situation in which the value of knowledge decreases. This happens when an information overload is generated. Some 'solutions' to this problem are: create improved management theories that aid selection among valuable and invaluable knowledge, create new learning norms that improve the use of the knowledge base, and the removal of obsolete management theories.
When dynamics increases, the storage of knowledge is useful for creating continuity, but can also act as a brake on innovations. Stored knowledge can be used for learning from the past. A more interesting opportunity is that of innovation by connecting different core competencies. This requires a matrix organization structure or task force, because core competencies are generally not shared among departments or SBU, and an inspirational leadership that stimulates ideas and activities in the organization for connecting competencies that are seemingly very different and unconnected.
These considerations are generalized in table 5.16 that summarizes the use of conserved knowledge in lean and classic machine bureaucracies.
MB: Storage:
Lean Classic
Acquisition Much knowledge and data Much knowledge and data Retention Less, because much removal when needed
Closely connected with mind and motivation
Much
In archives, formal rules and procedures Retrieval Much. Applied to problem solving Less. Connected with procedures,
indirectly linked with problems
Table 5.16: Differences in Knowledge Storage Between Lean and Classic Machine Bureaucracies
Use of knowledge
Ideal typical machine bureaucracies use knowledge developed in the past. It is therefore very conservative, but one could also think that change is not necessary because of the stability of the environment. Dynamic environments do require changes of knowledge, which is hampered by reification processes resulting from tradition. Using stored knowledge can also lead to competency traps, as was mentioned in chapter 4 (Kim, 1993; Levitt and March, 1988). Because the environment is simple, much knowledge is tacit. Making it explicit is sometimes very difficult but essential for reliable reapplicability of the knowledge.
The formal functional organization of machine bureaucracies leads to a very specialized use of knowledge, and even to knowledge ownership. This is not only the result of a political constellation that exists in machine bureaucracies, but also of problems in applying knowledge created elsewhere. The problem of the applicability of knowledge is linked with the fact that departments often lack a shared body of knowledge.
Dissemination of knowledge
problematic in complex situations with low codification: • Distribution of messages. This is the physical process.
• Mutual understanding. This is the semantic aspect of message dissimenation. • Synchronization. This implies that people's understanding matches in time. The first issues are standard problems mentioned in the communication literature (Stamper, 1973; Guetzkow, 1965). The synchronization issue is less well treated, but essential, because people must act in concert. When one organization member is still busy selling product X while other members have already found out that selling X only leads to losses and thus must be stopped, the organization is acting inconsistently because of the lack of a shared body of knowledge. This synchronization issue is well treated in the database literature (Rochfeld and Tardieu, 1986).
In stable environments, knowledge dissemination consists of regular reports and formal data streams. Low complexity environments can more easily create unambiguous information. Everybody receives precisely the data/information needed for his particular job and a precise data distribution schedule exists restricting synchronization problems. When environmental characteristics are very stable, speeding up knowledge dissemination by automation can be very effective. When complexity increases, standard reports no longer suffice. Dispersed knowledge must be connected to find solutions for complex problems (task groups) increasing the risk of asynchronous communication. Media richness should be increased to lower ambiguity and increase understanding (Daft and Weick, 1986). High dynamics demands faster communication channels too, and some delegation of responsibilities.
Adaptation of knowledge
Low complexity and a rather stable environment lead to slow and incremental changes in organizational knowledge. Actions in this situation are strongly motivated by action plans that have undebated models containing means-goals theories as their foundation. Not using these models can lead to actions that are not-legitimated. This can lead to severe sanctions when with hindsight these actions seem to have been ineffective. Adaptation of knowledge is mostly done by the person responsible for the knowledge. In the highly differentiated structure of machine bureaucracies this implies that knowledge adaptation is a specialist (technostructure) activity. Because feedback about mistakes is a slow process, adaptation is a time-consuming business, often leading to the implementation of knowledge that is already invalid. Additionally, the low risk of the environment demotivates organization members to start adaptation processes.
Performance control for evaluation and knowledge adaptation is often not done and not enough attention is paid to it. For instance in the Dutch high-tech company mentioned in chapter 4, at an assembly unit in 1992 production norms were used
that had been developed in 1968. In this organization (with fast moving dynamics because of rapid products and process innovations, this led to norms that are not applicable for the effective steering of the units. Strangely enough the organization has not changed its production norms in 14 years. Now that the company has come into a very hostile and competitive environment, it is being forced to reconsider its norms. This study is not being carried out by its own technostructure or management, but by one of my M.Sc. students! The rules, norms and procedures have reified the organization in such a way that it is not capable even of realizing a single-loop process.
The adaptation process can provide important triggers for double-loop learning. This happens when the complexity of the environment increases, so that the management theory cannot give a valid explanation or offer effective proposals, or when the dynamics has increased to such an extent that a theory must be found that improves understanding. Often uncertainty avoidance strategies are regarded as more effective than learning. In the longer run this could be untrue, as was shown in the case of lean production.
Definition of a Single-Loop Learning Score
The rating of learning activities can be done by noting the number of learning fields an organization is concerned with, and the number of activities undertaken as was done for the double-loop learning score. A further description of a scale for single- loop learning is presented in chapter 7.