One of the important uses of communication, in organizations and elsewhere, is for teaching and learning. An organization’s knowledge comprises the (relevant) knowledge stored in the memories of its members together with the knowledge stored in its files and records, including, nowadays, the data banks in its computers. Organizational learning is the set of processes that lead to the acquisition of this knowledge. Both employees and computers can participate in instructional processes either as teachers or learners. (A venerable example of a non-human teacher is a book; a more contemporary example is an intelligent computer tutor.)
The boundary between one biological organism and others is defined by identity of the shared DNA of all the organism’s cells. In a similar way, one might say that shared information determines the boundary of an organization—although the sharing is not nearly as complete as it is among an organism’s cells. Understanding the processes of organizational learning is critical to understanding the respective roles of organizations and markets in the economy. Shared knowledge makes it possible for organizations to behave in effectively coordinated ways that are not as easily available to coteries of independent firms.
Organizations acquire knowledge in the forms both of facts and pro- cedures. Much of the knowledge contained in human memories and machines resides in programs that govern the day-to-day activities of organization members and information processors. These procedures affect not only the behavior of individual employees, but also their relations with each other.
1. The Individual and Organizational Levels
The first question one must raise is whether organizational learning is different from learning by individuals. A recruiter who is interviewing a job prospect is learning about the candidate, and on the basis of that learning, together with other information, will or won’t make a job offer. As this learning by an individual has consequences for an organizational decision, providing new facts about the qualifications of the candidate, it must count as organizational learning.
If we adopt too strict a definition of organizational learning, we will define the topic out of existence. All human learning takes place inside individual human heads; an organization learns in only three ways: (a) by the learning of its members, (b) by ingesting new members who have knowledge the organization didn’t previously have, (c) by introducing new knowledge into its files and computing systems. For the moment, I will limit the discussion to human learning; learning by computers will be considered later.
What is stored in any one head in an organization may not be unrelated to what is stored in other heads; and the relation between those two (and other) stores will have a great bearing on how the organization operates. What an individual learns in an organization is very much dependent on what is already known to (or believed by) other members and what kinds of information are present in the organizational environment. As we shall see, an important component of organizational learning is internal learning—that is, transmission of information from one organizational member or group of members to another. Individual learning in organizations is very much a social, not a solitary, phenomenon.
However, we must also be careful about reifying the organization when talking about it as “knowing” something or “learning” something. It is usually important to specify where in the organization particular knowledge is stored, or who has learned it. Depending on its actual locus, knowledge may or may not be available at the decision points where it would be relevant. Since what has been learned is stored in individual heads (or in files or data banks), its transience or permanence depends on what people leave behind them when they depart from an organization or move from one position to another. It also depends on what records remain readable when computer software is changed. Has the knowledge been transmitted to others or stored in ways that will permit it to be recovered when relevant? Human learning in the context of an organization is very much influenced by the organization, has conse- quences for the organization, and produces phenomena at the organizational level that go beyond anything we could infer simply by observing learning processes in isolated individuals.
Let me perseverate for a moment on that term “organizational level.” Readers of March and Simon’s Organizations7 have sometimes complained that it was not a book on organizations at all but on the social psychology of people living in an organizational environment. The complaint was usually registered by sociologists, and was not without merit. We need an organization theory because some phenomena are more conveniently described in terms of organizations and parts of organizations than in terms of the individual human beings who inhabit those parts.
There is nothing more surprising in the existence of those phenomena than in the existence of phenomena that make it convenient for chemists to speak about molecules rather than quarks. Employing a more aggregate level of discourse is not a declaration of philosophical anti- reductionism, but simply a recognition that most natural systems do have hierarchical structure, and that it is often possible to say a great deal about aggregate components without specifying the details of activity within these components.
Hence, the remarks that follow have little or nothing to say about the detailed mechanisms that enable an individual human being to learn, but focus on how information is acquired by organizations, is stored in them, and is transmitted from one part of an organization to another. They are concerned with what are usually called emergent phenomena at the organizational level.
2. The Structure of Roles
For purposes of discussing organization learning, organizations are best viewed as systems of interrelated roles. As has been explained in the commentary on Chapter VI, a role is not a system of prescribed behaviors but a system of prescribed decision premises. Roles tell organization members how to reason about the problems and decisions that face them: where to look for appropriate and legitimate informational premises and goal (evaluative) premises, and what techniques to use in processing these premises. The fact that behavior is structured in roles says nothing, one way or the other, about how flexible or inflexible it is.
Each of the roles in an organization presumes the appropriate enactment of the other roles that surround it and interact with it. Thus, the organization is a role system.
3. Organizational Learning and Innovation
Since the organizations I know best are universities, I will draw upon my university experiences for most of my examples of organizational learning phenomena. Consider a university that wants to innovate along some dimension of educational practice—perhaps by building its instruction around the Great Books, or by focusing on something it calls liberal-professional education. I’ll use the latter example, which is closer to home.
The graduate schools from which a university draws its new teachers are organized in disciplines, some of which are saturated with the values
the best of my knowledge, that fly the banner of “liberal-professional” education. Clearly, a university that wishes to implement this view of instruction is faced with a major learning problem for its new (and probably its old) faculty members. It has no chance of accomplishing its goal without substantial education, and re-education, of its inductees. Moreover, the re- education is not a one-time task but a continuing one, unless the educational climate of the environing society changes so that it begins to produce graduates already indoctrinated with the desired goals and information.
Effects of Turnover. Turnover in organizations is sometimes considered a process that facilitates organizational innovation—getting out of the current rut. But in the case before us, where the organization is trying to distance itself from general social norms, turnover becomes a barrier to this kind of innovation, because it increases training (socialization) costs. To preserve its distinct culture, an organization of this kind may try to train its own personnel from the ground up, instead of relying on outside institutions to provide that training. Such inbreeding will have other organizational consequences.
Contrast this with the organization that finds in its environment training organizations that share a common culture with it. The Forest Service, in Herbert Kaufman’s classical account of it, is such an organization, counting on schools of forestry to provide it with new employees who are already indoctrinated with its values and even its standard operating procedures.8 The same thing occurs, less specifically but on a larger scale, in such professions as engineering, where there are close links between the engineering colleges and the industries, with a feedback of influence from industry to the engineering curricula.
An Experiment on Stability. If turnover is sufficiently low, organizational values and practices can be stabilized by the fact that each new inductee finds himself or herself confronted with a social system that is already well established and prepared to mold newcomers to its procedures. This phe- nomenon can be produced in the laboratory (and I believe actually has been produced, but I cannot put my hands on the appropriate reference).
In a certain experimental paradigm in social psychology (often called the Bavelas communication network) different patterns of communication are imposed on five-person groups. In one pattern (the wheel) one member of the group serves as leader or coordinator and all the other members communicate with him or her, and not directly with each other. In another pattern (the circle) the members are arranged in a symmetric circular network, each member communicating only with the two who are immediately adjacent. The groups are performing a task that requires them to share information that is given to the members individually.9
Now consider two groups whose members are Al, A2, A3, A4, A5, and Bl, B2, B3, B4, B5, respectively, where the A’s are in the wheel pattern and the B’s in the circle pattern. After they are thoroughly trained in the task, we open all the communication channels so that each member can communicate directly with all the others in that group. If they are under sufficient pressure to perform rapidly, the first group will likely continue to use the wheel pattern of communication and the second group the circle pattern.
After a number of additional trials, interchange Al and Bl. One would predict that the groups would continue to use their respective patterns. After a few more trials, interchange A2 with B2, then A3 with B3, and so on until the original wheel group is populated by Bl through B5, and the original circle group by Al through A5. We would predict that the A’s would now be communicating in a circle pattern and the B’s in a wheel pattern. If the experiment works as predicted, it demonstrates an emergent property of an organization—a persistence of pattern that survives a complete replacement of the individuals who enact the pattern.
The Problem of Sustaining Distinctiveness. The example of the deviant uni- versity can be extended to virtually all organizational innovation. Among the costs of being first—whether in products, in methods of marketing, in organizational procedures, or what not—are the costs of instilling in members of the organization the knowledge, beliefs, and values that are necessary for implementing the new goals. And these costs can be exceedingly large (as they are in the case of a university). The tasks of management are quite different in organizations that can recruit employees who are prefashioned, so to speak, than they are in organizations that wish to create and maintain, along some dimensions, idiosyncratic subcultures.
The mechanisms that can enable an organization to deviate from the culture in which it is embedded are, therefore, a major topic in organizational learning. As my university example suggests, this topic can be examined in the field, and particularly in a historical vein, by following the course of events in organizations that seek to distance themselves along one or more dimensions from the surrounding culture.
4. Organizational Memory
Retaining the unique traits of an organization is a part of the more general phenomena of organizational memory. Because much of the memory of organizations is stored in human heads, and only a little of it in procedures put down on paper (or held in computer memories), turnover of personnel is a great enemy of long-term organizational memory. This natural erosion of memory with time has, of course, both its advantages and disadvantages. In the previous section I emphasized one of its disadvantages. Its advantage is that it automatically removes outdated irrelevancies (but without discriminating between the relevant and the irrelevant).
Turning from the erosion problem, how are we to characterize an organization’s memories? Research in cognitive psychology in recent years has made great progress in understanding human expertise, and what has been learned was summarized in the commentary on Chapter V. The knowledge of experts is stored in the form of an indexed encyclopedia, which is technically referred to as a production system, so that whenever appropriate cues are evoked by a stimulus, access is provided to the corresponding chunk in semantic memory. Armed with knowledge stored in his or her production system, the expert is prepared (but only in the domain of expertise) to respond to many situations “intuitively”—that is, by recognizing the situation and evoking an appropriate response—and also to draw on the stored productions for more protracted and systematic analysis of difficult problems.
Against the background of this picture of expertise, the memories of an organization can be represented as a vast collection of production systems. This representation becomes much more than a metaphor as we see more and more examples of human expertise captured in automated expert systems. One motive for such automation, but certainly not the only one, is that it makes organizational memory less vulnerable to personnel turnover.
5. Ingesting Innovations from Without
My previous example had to do with organizations trying to retain their identities in a world of alien ideas, fighting the threat of increasing entropy that comes with the absorption of new personnel. The other side of the coin is the problem of assimilating innovations that originate outside the organization, or that have to be transmitted from a point of origin in the organization to points of implementation.
Research as a Learning Mechanism. So-called research universities have a dual mission: to create new knowledge and to transmit that knowledge to their students. Research accomplishes the former, and instruction the latter. Of course the real pattern is much more complicated than this. In the first place, the new knowledge produced by research is usually not initially transmitted only to students at the same university, but to researchers throughout the world, mainly by publication. In the second place, most of the knowledge transmitted to students in a university is not produced at that university. Is there really any reason why the research (which is one process of learning) and the instruction (another learning process) should go on in the same institution?
When we examine the research process more closely, we see that it differs rather fundamentally from the usual description. In any given research laboratory, only a tiny fraction of the new knowledge acquired by the research staff is knowledge created by that laboratory; most of it is knowledge created by research elsewhere. We can think of a research scientist as a person who keeps one eye on Nature and the other on the literature of his or her field. And in most laboratories, probably all laboratories, much more information comes in through ‘ the eye that is scanning the journals than the eye that is looking through the microscope.
It is probably true, and certainly widely suspected, that in any field of research a large fraction of the less distinguished laboratories could vanish without seriously reducing the rate at which new knowledge is created. Does that mean that these dispensable laboratories (dispensable in terms of the creation of knowledge) do not pay their way? The conclusion does not follow if the main function of a laboratory is not the creation of knowledge but the acquisition of knowledge. In military parlance, we would label such laboratories intelligence units rather than research units. They are units of the organization that are specialized for the function of learning from the outside world (and perhaps sometimes creating new knowledge themselves).
As a matter of fact, in universities we sometimes recognize the intelli- gence function of “research.” When we are asked why we require faculty members who are primarily teachers to publish in order to gain promotion or tenure, we answer that if they do not do research, they will not remain intellectually alive. Their teaching will not keep up with the progress of their disciplines. It is not their research products that we value, but their engagement in research which guarantees their attention to the new knowledge being produced elsewhere.
R&D and Manufacturing. The problem of developing new products from (local or imported) research ideas and of carrying them to the stage of successful manufacture and marketing is a classical organizational problem of creating and transferring information. It has already been discussed briefly in the commentary on Chapter II and will receive further attention in the commentary on Chapter XI.
In whichever direction the ideas flow through the organization, it is clear that nothing will happen unless they do flow. Normally, the learning associated with a new product must be highly diffused through the organization—many people have to learn many things—and such lateral diffusion and transfer is far from automatic or easy. It must overcome motivation obstacles ( I have already mentioned the NIH syndrome), and it must cross cognitive boundaries.
Manufacturing Constraints. A common complaint about contemporary American practice in new product design is that the design process is carried quite far before manufacturing expertise is brought to bear on it. But ease and cheapness of manufacture can be a key to the prospects of a product in competitive markets, and failure to consider manufacturability at an early stage usually causes extensive redesign with a corresponding increase in the time interval from initial idea to a manufactured product. These time delays are thought to be a major factor in the poor showing of many American industries in competing with the Japanese.
We know some, if not all, of the conditions for making communications between designers and manufacturing engineers effective. Each group must respect the expertise of the other, and must acknowledge the relevance of that expertise to their own problems. Moreover, each must have a sufficient knowledge and understanding of the others’ problems to be able to communicate effectively about them. Experience shows that these conditions are unlikely to be satisfied unless members of each group (or a sufficient number of members of each group) have had actual experience with the activities and responsibilities of the other group. In typical Japanese manufacturing practice, this shared understanding and ability to communicate is brought about by extensive lateral transfer of engineers in the course of their careers.
6. Acquiring New Problem Representations
In my earlier discussion of a culturally deviant organization, I contrasted the way in which roles (decision premises) are acquired in such an organization with the way in which they are acquired in an organization that builds upon the culture of the society that provides it with new members. Learning may bring new knowledge to bear within an existing culture and learning may change the culture itself in fundamental ways. I would like to turn now to that distinction.
In the past thirty years, a great deal has been learned about how people solve problems by searching selectively through a problem space defined by a particular problem representation. Much less has been learned about how people acquire a representation for dealing with a new problem—one they haven’t previously encountered.
Two cases must be distinguished: (1) The learner is presented with an appropriate problem representation, and has to learn how to use it effectively. That is essentially what is involved when organizations, already formed, ingest new members from an alien culture. (2) The organization is faced with a totally new situation, and must create a problem representation to deal with it, then enable its members to acquire skill in using that repre- sentation. In the extreme case, a new organization is created to deal with a new task. A new problem representation and a role system are created.
Creating an Organization. Some years ago I was fortunate enough to have a grandstand seat at the creation of the Economic Cooperation Administration, the U.S. governmental organization that administered the Marshall Plan of aid to Western European countries. In that process, which extended through most of the year 1948, competing problem representa- tions emerged from the very first days, each implying a quite different organization structure and set of organizational roles from the others. These problem representations were not made out of whole cloth, but arose from analogies between the presumed task of the ECA and other tasks that were familiar to the inventors of the representations from their previous training and experience.
For example, some participants in the planning drew an analogy between the ECA and wartime organizations that had supplied essential goods to the allies. Others thought of it as an exercise in investment banking. Others were reminded of the theory of international trade balances. From each of these views, a set of organizational roles could be recruited their new members. The commentary to Chapter XI will recount how this competition was resolved.
Why Representation Matters. Attention to the limits of human rationality helps us to understand why representation is important, and how policy may imply a representation. About two decades ago, the U.S. Steel Corporation began to contract its steel operations and to invest a major part of its capital in the oil industry, becoming USX in the process. The motivation of these moves was a particular representation of the corporation’s purposes.
If, a few years ago, you had asked executives of U.S. Steel what the corporation’s goals were, they might have answered: “To manufacture and market steel efficiently and profitably.” If you had persisted further, they might even have agreed that profit was the “bottom line.” But it would have been hard or impossible for them to describe the company without strong emphasis on its focus on steel. Their views might have been paraphrased: “We are out to make profits, but the way for us to make profits is to be an efficient steel manufacturer. That is a domain in which we have knowledge and expertise, and in which we can make good decisions.”
For the conglomerate that it became, an entirely different representation was required. The corporation has product divisions that can still be described in ways that resemble the earlier corporation—the word “steel” applying to some divisions, and “oil” to others. But in the new representation, these divisions are only components operating within a larger framework in which the fundamental policy is to invest available funds in the directions that will yield the greatest returns. Within that framework, new expertise is required: essentially the expertise of an investment banker.
Change in representation implies fundamental change in organizational knowledge and skills. It should not be surprising that under these conditions we often see massive turnover of personnel at all levels. It is often cheaper and quicker to import the new expertise and dismiss the old than to engage in massive re- education.
Source: Simon Herbert A. (1997), Administrative Behavior, Free Press; Subsequent edition.