Adaptive Models of Organization for Substance Abuse Treatment

Part I & II
July, 1998

Introduction

Part I: Organizational Adaptation
     - The Adaptive Systems Framework
     - Orientations Toward Use of Feedback
     - Applying the Adptive Systems Framework

Part II: Studying Organizational Effectivness And Adaptation
     - Levels of Analysis: Micro variables and Macro Patterns
     - A Systems Theoretic Approach
     - Transformational Dynamics in Organizations
     - The Burke-Litwin Model of Individual and Organizational Performance
     - Organizational Learning
     - Conclusions

Part III: Adaptation for Substance Abuse Treatment Organizations
     - Factors Affecting Change in Substance Abuse Treatment Organizations
     - Implications
     - Suggestions for Future Research
     - Conclusions
     - References

Introduction

This paper focuses on how to improve treatment through change, learning, development, innovation, and adaptation processes at the organizational level. It uses these approaches to suggest future directions of research and action for substance abuse treatment organizations.

Because of conceptual vagueness, the vast body of literature on organizational change and organizational learning frequently is unhelpful for guiding either research or practice. Part I attempts to clean up some of the conceptual clutter regarding organizational learning by presenting a series of simple models that highlight the four distinguishing characteristics of adaptive systems: (a) feedback, (b) coordination, (c) reactive learning, and (d) generative learning. The framework presented is descriptive, so it does not dictate what should be done to improve the study or operation of organizations. Rather, it provides a definition of adaptation and a conceptually precise way to categorize certain phenomena related to organizational learning.

Part II provides a discussion of prescriptive approaches to understanding organizational change. The Burke-Litwin model, which provides a complex causal model of organizational functioning, is used to guide organizational diagnosis and change efforts. The work of organizational learning theorists like Peter Senge (1990) is reviewed as a pragmatic approach for building a "learning organization." Part II concludes by integrating the literatures on organizational change and learning using the conceptual framework outlined in Part I.

Finally, in Part III, the theories and concepts discussed in Parts I and II are applied specifically to substance abuse treatment organizations. Through a series of theoretical propositions derived from the adaptive systems model explained in Part I, and to some extent in Part II, advice is presented for practitioners and social scientists. Social scientists can use the propositions as a starting point for conducting applied research by translating relevant propositions into empirical hypotheses. Practitioners can use the same propositions to develop actionable guidelines for how substance abuse treatment organizations can function more effectively. The paper concludes by suggesting some promising directions for future theory and research.

Sections

Part I: Organizational Adaption

Organizational learning and adaptation are relatively new areas of study in the social and behavioral sciences. This section defines organizational learning and organizational adaptation and explains the essential processes by which they occur. A conceptual framework is used to present the nature of systemic adaptation. An application of this framework and some limitations of the framework also are discussed. In subsequent sections, the adaptive systems framework is used as a starting point to explore in more detail the nature of organizations that learn and the critical variables that govern organizational adaptation and change.

Sections

THE ADAPTIVE SYSTEMS FRAMEWORK

A fruitful discussion of organizational adaptation in general, and the application of these ideas to substance abuse treatment organizations in particular, first requires an understanding of what is meant by adaptation and what distinguishes adaptation from other sorts of organizational activities and functioning. This section lays out a conceptual framework based on four distinguishing features of adaptation in individuals and organizations: (a) discrepancy reduction through simple self-regulation, (b) flexible functioning through complex self-regulation, (c) learning by reactive self-improvement and problem solving, and (d) adaptive learning (see Table 1).

Each of these four adaptive features can be thought of as stages that build on each other. The corresponding models represent the five levels at which organizations may function, with none of the adaptive features just mentioned exhibited at level 1 and all four features at level 5.

Table 1. Models of organization based on five levels of organizational functioning. Each level of functioning depends on the number of adaptive features represented in the model (the level 1 model has none of the features; the level 5 model has all four features). The basis for organizational change is described for each level.

The simplest model of an organization incorporates none of the distinguishing features of adaptation. Figure 1 shows a generic input-throughput-output model of organization. The model is useful because of its simplicity: The organization exists to transform inputs into outputs. That is, value-adding throughput activities transform inputs into more valuable outputs.

Organizational Effectiveness

The effectiveness of an organization may be seen as either impact (magnitude of output) or efficiency (ratio of output to input). The trade-off between impact and efficiency is an important concept. A for-profit company generally regards efficiency as the ultimate goal. The profit equation P = R - E (Profit is equal to revenue minus expenses) may be equated with VS = O - I (Value Surplus is equal to the total value of output minus the total value of inputs used to produce them). Only through a value surplus may a private company sustain itself as a profitable enterprise. Similarly, only through a value surplus, albeit more broadly conceived (i.e., including intangible societal benefits), can a not-for-profit or subsidized endeavor justify its existence. This can be viewed as a positive feedback loop whereby success (i.e., a favorable ratio of outputs over inputs) secures a value surplus that can be used not only to sustain the organization, but to enrich, expand, and enlarge it.

Figure 2 shows an input-throughput-output model with the addition of a negative feedback loop. Feedback allows the system to strive for equilibrium. For example, a heater that is a level 1 system will continue to produce heat at the same rate regardless of whether additional heat is needed. It cannot itself assess when it is appropriate to turn itself on or off. At level 2, however, the heater becomes a thermostat capable of self-regulation. Thus, whereas a level 1 system has an open loop where the basic activity of the system remains constant and unrestrained, self-regulatory systems alter the rate of activity in order to achieve a steady state. There are two key components to the feedback feature of level 2 functioning: (a) self-regulation and (b) feedforward.

Feedback and Self-Regulation

A thermostat regulates temperature not by producing heat at a constant rate, but by attempting to minimize the gap between the actual temperature and the ideal temperature (see Figure 3). This model represents the cycle of monitoring (monitoring the ambient temperature), comparative evaluation (comparing the actual temperature with some predetermined target point or range of temperatures), and goal execution (altering heat output in order to reduce the ideal-actual discrepancy) of self-regulation as described in cybernetics (Weiner, 1948, 1961) and control theory (Carver & Scheier, 1981; Karoly, 1993; Powers, 1973a, 1973b, 1978). Miller, Galanter, and Pribram (1960) call this a TOTE (Test-Operate-Test-Exit) cycle (see Figure 4). In a TOTE cycle, a testing mechanism looks for a discrepancy and initiates the operate mechanism (the throughput) to eliminate the discrepancy. Discrepancy is tested again and, if the operate function has brought the present state to a satisfactory level, the testing mechanism terminates the process and exits.

Feedback and Feedforward

Implicit in the thermostat analogy is the goal of the system, which feeds forward the set point, the ideal temperature toward which the system strives. This feedforward, what Powers (1973a) calls a reference signal, is the basis on which self-regulating systems assess and correct discrepancies. The feedback mechanism of a thermostat measures the actual temperature, and through a process of comparison and reaction, the heating mechanism responds appropriately to reduce any gaps between actual and ideal temperature (i.e., the performance-standard mismatch) and maintain the steady state. This comparison-and-reaction process is regulated by the test mechanism of a TOTE cycle, also called a comparator because it compares existing conditions or performance with some preset standard toward which the system strives (see Figure 4).

Dynamic Steady States

Cannon (1929, 1932) and Lewin (1938, 1951) proposed the concepts of dynamic homeostasis and quasi-static equilibrium, respectively, as processes by which biological (Cannon) and psychological (Lewin) steady states are maintained. Because of constant environmental disturbances and fluctuations, the system oscillates back and forth around its target, but never precisely reaches it for any length of time. This is simply an acknowledgment that external conditions change—if they did not, self-regulation would serve no purpose.

One important aspect of negative feedback loops in self-regulating systems is time. More specifically, there may be a delay between the operate or throughput function and its impact on the test or monitoring function. A control system that does not take such a delay into account may behave inadvertently as a discrepancy-inducing cycle by continuously operating to close a gap that only gets larger due to a vicious cycle of overcompensation. For example, if a thermostat's thermometer reacted to changes in ambient temperature an hour after they actually occurred, then the heater would run for a full hour after the temperature had reached its set point. By that time, the temperature may have become so hot that the air conditioner would have been initiated, but it too would be expected to run for an hour longer than necessary due to the time delay. In some cases, such cycles can be self-amplifying, causing the system to stray farther and farther from the set point.

Feedback does not necessarily imply a changing relationship between the system and its environment. Although the environment might change, and the system may be active or inactive, the relationship between the system and the environment is fixed. This limitation is reflected in the nature of possible corrective responses to the performance-standard mismatch: Corrective response is limited to more or less persistent or intense use of a single fixed throughput activity.

Figure 5 incorporates the second feature necessary for adaptation: the regulation and coordination of multiple activities. The coordination of multiple activities interacting within the same system, along with the hierarchical feedback loops that engage them, creates an "elaborate and flexible functional capability" not present in level 2 systems with a single feedback loop (Ford, 1987). Carver and Scheier (1981) describe this as a system’s hierarchical decision-making chains, where the output of one feedback loop may serve as the feedforward for a subordinate loop. Thus, the feedforward for a thermostat is the output of a superordinate system—the human occupants who set the target temperature. For simplicity, Figure 5 shows this as the feedback loop leading from output back to output.

Thus, the output function displayed in Figure 5 represents not only output in the form of value or energy, but also information to help coordinate and control subordinate feedback systems, which, although not displayed as such, exist underneath the feedback system depicted here. These are called "nested" feedback loops.

The idea of multiple feedback loops implies that the system may now have multiple goals and that there may be multiple activities for achieving them. A computer-simulated chess opponent, for example, must consider a multitude of goals and subgoals. Even the terminal goal is not altogether clear; it can be to win the match or simply not to lose it. This will dictate tactics for, say, attaining or avoiding a stalemate. Both goals, winning and not losing, may be pursued simultaneously—successfully winning accomplishes both goals. However, these goals occasionally will conflict, such as when a draw is offered by the computer's opponent, in which case the computer must decide which is more important, to win or not to lose, and elaborate strategies must be employed to help it execute various activities and accomplish various goals.

Ford (1987) uses the example of a guided missile instead of a computer chess game. A guided missile has several options in terms of the activities it may perform: It can arm or disarm, go faster or slower, up or down, left or right; but it must execute these actions while attempting to both hit its target and avoid things that are not its target. Again, there may be a trade-off. The system, in pursuing multiple objectives simultaneously and negotiating the options presented, must coordinate activities in a much more complex fashion than, say, a thermostat. Where the thermostat had one simple rule, the guided missile must have meta-rules to govern and coordinate a complex network of rules. One meta-rule might be that the system gives precedence to the nontarget avoidance objective over the target pursuit objective under certain conditions (e.g., when it is armed rather than disarmed, when it is over civilian populations rather than uninhabited areas). These meta-rules, along with the stored behavioral routines or strategies used to execute them, constitute the regulatory and control functions of a system.

To be effective, a guided missile must simultaneously pursue its target and avoid nontargets. It is not enough to switch off between objectives and pursue one while neglecting the other. Actions must be coordinated carefully—the missile must be able to take evasive action while simultaneously minimizing deviation from its target. This sort of coordinated execution of multiple objectives results in a much more elaborate system than that presented in the previous model. In level 3 systems, multiple feedback loops and hierarchical rule structures may be used to create very smart missiles and computer chess players.

Although complex control systems may be thought of as "smart" in some ways, they are also very dumb: They still cannot learn. Their behavior is limited to exactly what their designers have programmed. Although the rules, meta-rules, and functional strategies may be intricate, the system is just a fancier version of the thermostat. It may not modify itself or change its behavior to deal with novel conditions or unexpected problems in the environment.

Self-improving systems (see Figure 6) are expected to learn from their interactions with the environment. This is what Ford (1987) calls "adaptation through self-programming." The system not only may now change which activities it executes and to what degree, it may change and improve the nature of the activities themselves. Whereas level 3 functioning may achieve improvements in efficiency and effectiveness through better coordination of activities and allocation of resources, the activities that are coordinated and those to which resources are allocated are still fixed. Level 4 functioning represents an improvement of the activities themselves. At level 4, feedback is used not only to regulate, but to learn.

The Learning Function of Feedback

The feedback loop now serves not only to control inputs, but also to learn and therefore improve the throughput activities. David Nadler (1977) pointed out two implications of this learning function of feedback. First, feedback now can be used both to regulate the consumption of input and intensity/persistence of output and to improve the throughput activities themselves. "Feedback can be more than just a device to regulate flows through the system—it can be a way to change how the system does its work" (p. 69). Second, due to the increased complexity of the model, feedback now comes in many forms and may be used, or not, in different ways. The system now may seek out different sorts of feedback relevant to potentially new throughput activities. System dynamics may be considerably more complicated because feedback, standard-matching, and self-correction routines are no longer straightforward and automatic. Relevant information may exist, but the system may choose not to use it.

Nadler’s second point involves not only the choice of whether or not to feed back relevant information, but also the issue of where, and the extent to which, information and feedback is sought. In other words, what is the intensity and direction of the system’s attention? A system can direct its attentional resources in a number of ways. One simple but important dichotomy is based on the source of the information—internal or external—to which the system attends. If the source of information is internal, the system is attending to itself (self-focus), whereas external information means that the system is attending to its environment (environment focus). Psychologists have tended to assume that individuals’ attentional resources are fixed (e.g., Carver & Scheier, 1981; Duval & Wicklund, 1972; Kanfer & Ackerman, 1989; Kanfer, Ackerman, Murtha, Dugdale, & Leissa, 1994). This implies that, especially when attentional resources are taxed, attending more to the environment will necessitate attending less to the self, and vice versa. Thus, attention will tend to shift between being primarily self-directed and primarily environment-directed, depending on various aspects of the person and the situation (Carver & Scheier, 1981).

Whereas the attentional resources of individuals may be fixed, an organization’s ability to attend to itself and/or its environment is more likely limited by the value surplus created by its throughput activities. The more energic inputs it can secure relative to the amount used by its operate, or throughput, function, the more energy that is available to the regulatory activities of the system such as attending to relevant internal and external information. Where information relevant to a particular goal-activity relationship is not used, the organization likely functions as if it were an open loop system (i.e., a model one system lacking the feedback loop), and it will habitually act, or fail to act, in the same way over time. Only when internal or external conditions cause the organization to redirect its focus toward the appropriate information can the organization alter the relevant behaviors.

Level 4 systems are the first presented here that may be called adaptive. However, the potential of these systems is restricted to fixed terminal objectives. That is, although operational objectives or subgoals may be altered, or new ones invented (entire layers may even be added to the existing hierarchy of rules and feedback loops), they serve a set of fixed terminal goals. For instance, although efforts are currently underway to build a self-improving computer chess player that will learn and improve the more it plays, its terminal goals of winning and not losing are fixed. This computer will never throw a match in order to make its opponent feel better (even if the opponent was a sore loser and this behavior was prudent for the computer’s survival), nor will it ever get bored and decide to learn a new game. It is not self-aware, the hierarchy of goals and feedback loops is finite, and its terminal goals do not change. Although it may learn, and is therefore adaptive in a reactive sense, it does not have the adaptive qualities of living organisms and social organizations, which are adaptive in terms of activities and the terminal goals they serve.

The adaptive systems model (Figure 7) depicts systems as dynamic not only in terms of input-throughput-output activities, but also in terms of the goals or intended outcomes of those activities. Regulatory and control processes may use feedback in four ways: (a) to alter resource inputs via self-regulation; (b) to select among different means of achieving goals; (c) to alter means for achieving goals; and (d) to alter the basic goals or purposes toward which the system is striving.

Adaptive Systems Are Guided by Purpose

Whereas the temperature of a room, the weather, the flow of water in a river, and the orbit of planets around a star are all complex systems, they are not guided by conscious and unconscious goals as sentient organisms and social organizations are. All previous models were able to use mechanical or computer analogies. Generative learning applies only to adaptive systems such as humans and human organizations. Level 5 systems are self-constructing and self-organizing, guided by terminal goals and purposes that are, at least to some extent, of their own creation.

Norman and Shalice (1986) observed that patients with frontal lobe brain damage are no longer able to self-generate goals and intentions. Although they respond perfectly well to situational demands and the instructions of others, they are incapable of planning or engaging in nonautomatic behaviors. This suggests a separate, superordinate mental function to generate intention that is distinct from the mental functions that carry out those intentions (Kuhl, 1994). Self-generation of intentions, however, does not occur in a vacuum—it relies on feedback from the environment.

The goals of adaptive systems are interdependent and may be conflicting. Inevitably, certain goals take precedence over others. According to need theories, the goals toward which humans strive may be understood according to a limited and finite set of basic, fundamental needs. The point is that whereas humans and social organizations may follow certain patterns in the formulation and pursuit of goals, and those patterns are influenced by a dynamic interaction between person and environment, the goals of humans are self-created.

Sections

ORIENTATIONS TOWARD USE OF FEEDBACK

The use of feedback and its effects on behavior are largely determined by goals and individual orientations toward either (a) learning and improvement or (b) decision making and performance. The most useful framework for understanding such orientations in individuals was provided by Carol Dweck and her colleagues in their theory of goal orientation (Dweck & Leggett, 1988; Elliott & Dweck, 1988). An individual’s orientation toward the pursuit of goals and the use of feedback on relevant performance tends to focus on either learning or performance. In the face of initial failure to achieve goals (i.e., negative feedback), those with a learning orientation will concentrate on using that feedback to improve future performance, whereas those with a performance orientation will tend to allocate their resources elsewhere.

Although the concept of "the learning organization" (e.g., Senge, 1990) is now taking hold in business lexicon, application of its principles requires a philosophy of learning that is based on challenging assumptions, detecting and correcting errors, and striving for improvement in the long run rather than demonstration of performance in the short run (Ackoff, 1993). For example, in their discussion of the philosophical underpinnings of total quality management, Hackman and Wageman (1995) note that although measurement and feedback are key to the learning process, this process will be "rendered impotent" if a learning orientation is not adopted over the short-term performance orientation that tends to pervade most organizations.

It is important to note that a learning orientation need not be at the expense of a performance orientation. In a discussion of Dweck’s theory on individual differences in learning orientation, Button, Mathieu, and Zajac (1996) noted that concern for performance and concern for learning are more likely independent axes than opposite poles of a single continuum. Thus it is possible to be high on both. This is also true for organizations. After all, an organization that does not demonstrate requisite performance in the short run may not survive. Thus, although learning results from a focus on improving performance in the long run, it is not incompatible with concerns for short-term performance as well.

Chaffee (1985) and Hart (1992) comment that the simultaneous adoption of multiple strategic orientations will have a greater impact on organizational competitiveness in the long run than a single orientation. Of the five orientations described by Hart, three are relevant in terms of our discussion: (a) the rational/performance mode, (b) the transactive/learning mode, and (c) the symbolic/purpose mode. In his discussion of strategic modes, Hart notes that:

Firms that are able to combine several modes into a high "process capacity" might be expected to perform [better] than single-mode or less process-capable organizations. For example, a firm that combines the elements of the symbolic and transactive modes would blend dedication to a shared vision and mission (symbolic) with a strong learning orientation (transactive). Such a firm should perform well in terms of growth, quality, and future positioning. (p. 345)

As the five "levels" described in the adaptive systems framework suggest, there is a hierarchical quality to these orientations. Etzioni (1967, 1986, 1989), for example, offers an organizational decision theory of "mixed scanning" where decision makers switch between various levels of the decision process, from high-level goal setting (like Hart’s symbolic/purpose mode) to subordinate goal execution processes (the rational/performance mode). Along the same lines, Chaffee (1985) notes that there is a hierarchical relationship between the three strategic modes, with the purposive or "interpretive" mode at the top, then the learning or "adaptive" mode, and finally the rational or "linear" performance mode.

The work of Etzioni (1967, 1986), and particularly of Hart (1992) and Chafee (1985), highlights a central theme of the adaptive systems framework: To be successful, organizations must effectively function at multiple levels simultaneously, including mechanistic or linear functioning (levels 1 through 3) to transactive learning (level 4) to generative learning (level 5).

Sections

APPLYING THE ADAPTIVE SYSTEMS FRAMEWORK

An example using synergy and team learning is presented below to clarify how the systems framework can be used as a lens through which to view other concepts related to organizational functioning and effectiveness. Some concluding comments are offered about the intended use of the framework.

Consider the concept of synergy through the perspective of the adaptive conceptual framework presented above. A useful tool for thinking about synergy is the "lost in the wilderness" exercise commonly used to stimulate group interaction and, it is hoped, synergy. In this exercise, participants are asked to imagine that they are stranded in some desolate location (e.g., the wilderness, the desert, an island, the Arctic) and must decide which items are needed for survival. The choice is typically constrained by either number or weight of the items. The list of items may include things like a compass, a frisbee, a knife, a rope, or a pencil. The idea is to select those items that will be most useful for your individual and/or group survival. For the sake of discussion, let us assume that there is a correct answer to this problem—some best combination of items that will maximize survival potential. To assess the idea of synergy, let us make the example more concrete with some assumptions. First, assume that there is a list of 40 items and your group can take only 20, so scores can range from 0 to 20 correct responses on the exercise. Also, assume there are four group members participating in the exercise: Anne, Bob, Chris, and Dorothy. If these members took the "survival" test individually, they would score as follows: A = 7, B = 2, C = 15, and D = 8. Finally, assume that among all their selections, only one of the 20 correct items (say, a screwdriver) would be chosen by none of them. Now, let us consider what the likely score of the group will be according to each model.

Level 1 systems have no feedback. There is no way to regulate flow through the system. Thus, each person would have equal contribution to the decision, but individual members would have to act independently, since this is a mechanistic model. If a voting or round-robin selection process ensued, so that all members were equally but individually operating to make a decision, we would expect the group to achieve the average of their individual scores. In this example, the group would achieve a score of 8.

Level 2 systems have feedback, but it is still an individual affair (coordination is the hallmark of level 3 systems). Thus, Bob may realize from past experience that he is terrible at this exercise, so he may decline to contribute. The remaining three group members would then operate in the same way already described, but now the group average consists of only three scores, so the group raises its average score from 8 to 10. By this logic, it is even possible for three of the members to refuse to contribute based on past experience so that Chris brings the group average up to his individual score of 15.

A complex system adds an important new component: coordination. Although this system may still only regulate flow through the various fixed activities of the system, functioning is considerably more flexible. Different members of the group may have different areas of expertise. By managing the process, the group may now allocate particular activities or choices to the people with relevant expertise. For instance, the nutrition expert might select the correct food, whereas the engineer could pick the best tools. In this way, the group can achieve cooperative synergy. The group can score 19 by properly coordinating and allocating the group’s resources. Of course, the downside is that, although the group could achieve a score of 19, it might also achieve a score of 2. Another aspect of level 3 systems is hierarchy, and if Bob is the "team leader" and forces his own personal selection based on formal position or persuasive power rather than expertise, the group will end up with his dismal individual score of 2. We might call this "negative synergy."

At level 4 functioning, the group may learn, and a new capability not possible among the members individually might be created through team synergy. At this level of functioning, the team could potentially, through discussion and pooling of knowledge, figure out that the screwdriver (the one item none would have chosen individually) is an important item to have, thus raising the score to 20. Let us not forget, however, that there also may be a downside to group process. As Asch (1952) and Janis (1972) have demonstrated, subtle pressures on individual members to conform to the will of the group, perceived or actual, can lead to disastrous blunders that few, if any, of the group members would have committed individually.

Functioning at level 5, the group would be able to learn how to function better as a group. They might achieve only a score of 16 or 17 (scores cannot reflect level 5 functioning), but by reflecting on the group process and engaging in generative learning, they would better understand themselves individually and in relation to one another and their environment. Through generative learning, the group would be able to question the purpose and utility of the exercise; to explore group processes and the functionality of their shared assumptions; and to improve collaboration, cohesion, and group functioning. Emotional relationships might even replace task accomplishment as the most important aspect of the group for its members. There are advantages and disadvantages to functioning at this level. The effectiveness of a task group with a clear goal and unambiguous work might suffer, for example, if time were unnecessarily spent reframing the problem at hand and questioning the purpose of the group. In this case, level 5 functioning would merely act as a distraction. On the other hand, in less straightforward circumstances, dealing with process issues within the group and reconsidering its purpose might be expected to enhance the long-range cohesion and success of the group and its members.

The adaptive systems framework is an attempt to clarify the nature and underlying processes of organizational change. However, the framework provides only a conceptual tool with which to think about organizations. An essential point of clarification now must be made regarding the five models presented above: The adaptive systems framework put forth in this paper is intended to be descriptive, not prescriptive. This framework offers a taxonomy for categorizing levels of systemic functioning, from static (level 1) all the way to complex and generative (level 5). The point is not that an individual, group, or organization gets stuck at a suboptimal level and needs to "get to level 5" to be effective. The framework is better understood in terms of the implicit theories people have about how a system functions. In truth, individuals and organizations necessarily function at all five levels. With few exceptions (e.g., frontal lobe patients), humans and organizations are always best described by the generative model, whether they think they are or not.

Functioning at the level of generative learning (level 5) is descriptive of how individuals and social organizations actually function. This level of functioning can be consistent or inconsistent, to a great or small extent, explicit or implicit, conscious or subconscious, systematic or unsystematic, effective or ineffective. By conceptualizing an organization as simple (model 2), mechanistic (model 3), or reactive (model 4), we inhibit that organization's conscious, consistent, and effective generative learning processes that are increasingly critical for the survival of modern organizations. Kuhn’s (1962) point that major scientific breakthroughs are often initiated by mistakes or unexpected findings that challenge our underlying assumptions or "paradigms" makes it clear that we need not consciously pursue generative learning for it to occur. The Hawthorne studies, for example, originally set out to complete a series of time and motion studies, but found an entirely unexpected phenomenon resulting in what Kuhn (1962) calls a paradigm shift. This does not, however, preclude us from conscious attempts to make mistakes, challenge assumptions, and consistently revisit our implicit understanding about the nature of problems, meaning, and purpose.

Although the framework presented here cannot do justice to the complexity of individuals, much less groups and organizations, it does tap the essence of the general patterns of interrelationship that are organizations. Thus, the adaptive systems framework is useful because (a) it defines and explains organizational change and adaptation in terms of four essential characteristics, and (b) it reflects the mental models people use when thinking about organizations. The first useful feature, describing the systemic characteristics of organizational change, provides a foundation on which to build more pragmatic causal models of organizational functioning. In Part II, we explore these prescriptive, pragmatic notions that more directly guide practitioners and applied social scientists.

Sections

Part II: Studying Organizational Effectiveness and Adaptation

Although systems theory and the adaptive systems framework presented above offer useful ways of thinking about organizations, they do not directly point to specific methods by which organizational functioning can be improved. Rather, their value lies in describing how the systems concept links individual behavior to organizational dynamics. This section begins by exploring how a systems approach, and the adaptive systems framework in particular, helps link micro variables at the individual level to broader patterns of organizational functioning. Two pragmatic theories of organizational change and learning based on the systems approach will then be presented, followed by an attempt to integrate the two theories based on the adaptive systems framework.

Sections

LEVELS OF ANALYSIS: MICRO VARIABLES AND MACRO PATTERNS

Katz and Kahn call the chicken and egg dilemma of specific micro variables and overarching macro patterns the "problem of levels" (1978, p. 12) and propose that a systems theoretic approach is the key to resolving the dilemma. They argue:

The weakness of the micro approach in the past has been. . . [1] It has dealt with too few of the significant variables in the total situation. [2] It has often seized upon inappropriate variables and has pushed too hard in the direction of showing the universality of some fundamental principle. (Katz & Kahn, 1978, p. 14)

Katz and Kahn conclude by noting that "this reductionistic emphasis in its very character tends to lose the problem with which it should be concerned" (p. 14).

Although sociologists have taken the first steps toward identifying key relationship patterns at macro levels of analysis (e.g., organizational and societal), they have failed to attain anything beyond a cursory understanding of these patterns because of their inability to link them to the underlying psychological and behavioral phenomena that must be measured to get a clear picture of the relationships with which they are concerned. Thus, sociologists have broadly identified some important patterns, but have not identified the measurable psychological and behavioral variables that, through their interrelationships, create those patterns.

Sections

A SYSTEMS THEORETIC APPROACH

Systems theory enables scholars to start at the point of macro patterns and break them down into component micro concepts without losing the link of the parts to the whole (c.f., Katz & Kahn, 1978; Mayntz, 1997). Thus, the psychological and behavioral dynamics that make up macro patterns can be selected and studied in relation to the pattern as a whole. The adaptive systems framework suggests that the variables critical to organizational development and learning are the governing variables of the system. Governing variables consist of the overarching purposes, directions, and strategies for the system as a whole, as well as the leadership and decision-making processes by which they are set. By linking the "micro" to the "macro" through governing variables, we are able to better understand the patterns of behavior that create and sustain organizations.

The adaptive systems framework is a conceptually abstract approach to understanding organizations. It does not define the specific variables that "govern" the system, nor does it state the nature of their patterns of interrelationship. Furthermore, the adaptive systems framework is purely descriptive. Although it may facilitate a conceptual understanding of systemic adaptation and learning, it does not offer concrete or practical suggestions about how organizations work or how to improve their effectiveness.

More practical approaches to studying organizations and improving their effectiveness come from two related streams of scholarly work. The first concerns an understanding of the transformational dynamics of organizations, concentrating primarily on a model proposed by W. Warner Burke and George Litwin. Their model of individual and organizational performance (Burke & Litwin, 1992) is not merely descriptive; it is also predictive and, by extension, prescriptive. The Burke-Litwin model offers one useful approach to conceptualizing and studying the governing variables of organizations, how they relate to other key organizational variables, and the dynamics of organizational change and development. The second stream of work revolves around organizational learning, including a review of Peter Senge’s work (1990) on the learning organization. Not surprisingly, the Burke-Litwin model of organizational effectiveness and Senge’s work on organizational learning both stem from a systems theoretic approach to understanding organizations.

Sections

TRANSFORMATIONAL DYNAMICS IN ORGANIZATIONS

The Burke-Litwin model (1992; Burke, 1994) is an ideal starting point for understanding the transformational aspects of organizational dynamics and organizational change. First, however, it is necessary to understand what transformational means, and the distinction between transformational and transactional dynamics. The distinction between transformational and transactional organizational dynamics is based on Burns’s (1978) conceptions of leadership. This section begins by summarizing the conceptual distinction between transformational and transactional leadership offered by Burns (1978) and others (Bass, 1985; Kanter, 1983; Rost, 1991; Tichy & Devanna, 1986).

The leaders and managers of an organization can be viewed in terms of the specific functions they serve or the behaviors in which they engage. According to Rost (1991), the dominant perspective on leadership in the 20th century has been leadership as effective management. Management and leadership are more appropriately viewed, however, as distinct processes. The distinctive nature of these processes can be understood using Burns's (1978) concepts of bureaucratic authority, transactional leadership, and transformational leadership.

The nature of bureaucratic management and administration is highlighted by the defining features of level 3 systems: hierarchy and coordination. Here, managers are seen primarily as Taylor and Weber viewed them—as coordinators, supervisors, and resource allocators. Bureaucratic managers rely on hierarchical authority, rules, and threat of punishment to execute their prescribed tasks and maintain organizational effectiveness. Burns (1978) and others (O'Toole, 1994; Rost, 1991) have pointed out that "leaders" who rule only by coercive power and legitimate authority are not leaders at all. Others have made similar distinctions between managers who use existing policies and structures to allocate resources and maintain organizational performance and leaders who initiate organizational change and development (e.g., Zaleznik, 1977).

Burns (1978) holds that the roles and behaviors of leaders and their followers are not fixed according to existing systems, policies, and structures, and are not based on coercive or authoritarian relationships. Instead, leaders are constantly redefining their roles and changing their behavior in an evolving exchange with followers. Burns's description of leaders "modifying their behavior as they meet responsiveness or resistance" (p. 440) is consistent with reactive learning in level 4 systems. Leaders do not simply rely on existing, formally prescribed duties and lines of authority to carry out their work. Instead, they may modify their behavior in order to more effectively achieve prescribed goals.

Transactional leadership processes are not concerned with altering the fundamental goals of the organization or its members. Thus, transactional leaders tend to influence and motivate through reciprocal exchange processes (c.f., Bass, 1985; Burns, 1978), appealing to the self-interest of others and rewarding desired behaviors quid pro quo. Although new behaviors may be learned, at both the individual and group levels, purpose in transactional leadership is fixed, consisting of the formally prescribed goals of the organization.

When discussing transformational leadership, Burns (1978) invokes terms like morally purposeful and the shaping of purpose, and Tichy and Devanna (1986) refer to planful opportunism and organizational self-renewal. The adaptive systems framework offers a distinction between transactional and transformational leadership that is perhaps more clear: Transformational leadership is based on processes of generative learning (level 5), whereas transactional leadership is characterized by reactive learning (level 4).

As with all adaptive systems, the purposes of organizations must evolve over time. The process by which this occurs for an organization and its members is transformational leadership. Rost (1991) summarizes this point by stating that the reciprocal influence relationship between leaders and followers is characterized by common purpose. Mutually shaping a common purpose over time suggests "development (progressive change) in the purposes of the leaders and followers rather than fixed, stable positions on what often are complicated and rapidly changing issues" (Rost, 1991, p. 120).

Transformational change is discontinuous, based on self-renewal of identity and purpose, whereas transactional change leaves the organization’s identity and goals fixed, instead focusing only on the means by which desired ends will be optimally achieved. The Burke-Litwin model takes advantage of this useful distinction by applying it not just to leadership, but to organizations in general.

Sections

THE BURKE-LITWIN MODEL OF INDIVIDUAL AND ORGANIZATIONAL PERFORMANCE

The Burke-Litwin model originates from the work of George Litwin and his colleagues (Litwin & Stringer, 1968; Tagiuri & Litwin, 1968) on organizational climate in relation to managerial behaviors, norms, and values; organizational structure and policies; and group and organizational performance (see Burke, 1994; Burke & Litwin, 1992). Litwin describes climate as the perceptions, expectations, and feelings of an organization or group. Climate is malleable, meaning it is often transitory and may be influenced both by enduring aspects of organizations (e.g., shared norms) and by transient aspects of organizations (e.g., supervisory behaviors). Litwin and Stringer (1968) experimentally manipulated work unit climate through managerial behavior and, based on their theory, predicted its impact on performance. This was an important first step toward building a predictive, not just descriptive, model of organizations.

The Burke-Litwin model (see Figure 8) is valuable not only because it specifies the causal relationships between key organization variables, but also because it clearly differentiates between transformational and transactional dynamics in organizations. Organization climate, for instance, is a transactional factor. Although it is strongly influenced by more enduring aspects of organizations, and therefore may be relatively predictable on average in the long run, it is also fairly easy to influence in the short term (Schneider et al., 1996). Changing the underlying, enduring aspects of organizations, such as culture, is considerably more difficult, but may have more impact on long-term performance because it operates at the transformational rather than transactional level of organization dynamics (Burke, 1994; Schneider et al., 1996).

Figure 8 presents a summary of the Burke-Litwin model. Following principles of a systems theoretic view (Katz & Kahn, 1978), the Burke-Litwin model begins with environment, from which inputs are brought into the organization, and ends with individual and organizational performance, or the outputs. In a sense, however, the model finally ends with the environment once again, as the model also includes a feedback loop linking outputs back to environment and system inputs. The factors depicted in between environment and performance are the throughput. It is here that the model offers some specificity on the mechanisms by which the organization does its work.

Transformational Dynamics

The transformational factors are represented by the top three boxes in the model presented in Figure 8. Consistent with our discussion thus far, Burke defines transformational factors as "areas in which alteration is likely caused by interaction with environmental forces (both within and without) and which require entirely new behavior sets" (1994, p. 129).

Mission and strategy refer to what the organization members hold to be the central purposes of the organization, and the general strategy by which the organization plans to pursue them. Note that this may or may not be accurately captured by a mission statement. This factor is based on what organization members believe to be the mission and the strategy, not necessarily what is written in a formal statement.

Burke (1994) describes culture as "the relatively enduring set of values and norms that underlie a social system. These may not be entirely conscious. Rather, they constitute a ‘meaning system’ that allows members of a social system to attribute meaning and value to the variety of external and internal events they experience" (p. 126). Culture is strongly influenced by the history, traditions, and customs that are passed down to new members over time.

Leadership is defined as the behaviors of senior management that establish direction for the organization and encourage organization members to take the requisite actions to move in that direction. Transformational leadership, as discussed above, and the values held and communicated by executives are included in this variable.

Leadership is central in the Burke-Litwin model, having a strong causal impact on culture and on mission. O’Toole’s (1994) notion of values-based leadership, for example, clearly establishes the link between culture and effective transformational leadership. Leadership is also closely tied to mission and strategy. Although formalized statements of mission and strategy represent the linear, analytical goals and strategies of the organization and its members, the purpose created by transformational leadership may occur through more creative direction-setting activities, such as creating a shared vision for the future that is vivid and inspiring to organization members. Tichy and Devanna (1986), for example, discuss how effective transformational leadership entails "right-brain" visioning before "left-brain" strategizing. Thus, the leadership-to-strategy arrow may require leaders to translate a creative, personal, emotional vision into a pragmatic, analytical mission and strategy. A vivid, well-communicated vision is an important part of charting a new direction for an organization and its members (Beckhard & Harris, 1987; Kotter, 1996; Tichy & Devanna, 1986).

The remaining throughput factors make up the transactional variables of the model. Changing transactional factors is easier than changing transformational factors, but will have less of a long-term impact on performance if transformational factors are unaffected and remain essentially unchanged.

• Structure includes the formal chain of command, prescribed lines of communication, responsibilities, and decision-making relationships.
• Systems are the standardized behavioral systems or ways of doing things, including the policies, procedures, and practices designed to facilitate the work of the organization’s members, such as reward systems and control processes (e.g., budgeting).
• Management practices are what managers do to use resources available to them to achieve organizational goals. These are the bureaucratic manager and transactional leader behaviors that are intended to pursue current purpose and execute existing strategy.
• Work unit climate is the combined sense of direction, perceived roles and responsibilities, commitment to and involvement in the organization, and perceived equity of rewards for a group’s or organization’s members. The climate of a work unit strongly affects intragroup and intergroup relations.
• Task requirements refers to the specific requirements of a given task, the set of tasks assigned to specific roles and job positions, and the skills and abilities of the individuals who fill those positions and complete the tasks. Ideally, jobs will be appropriately matched to the talents of the individuals expected to fill them.
• Individual needs and values represent the organizational factors that, at an individual level, fulfill important work-related psychosocial needs, such as a sense of autonomy and achievement, and a belief that the work—and the individual’s contribution to that work—is important and worthwhile.
• Motivation is defined by Burke as the "aroused" goal-directed behavior of the organization and its members. The enacted behavioral tendencies of the system stem directly from the energy generated by human motivation.

Because it is a causal model, the Burke-Litwin model can serve scholarly and empirical as well as practical purposes. Researchers may use the model to define important organizational variables, whereas practitioners can use the model to help guide change efforts. One key to using the Burke-Litwin model effectively is finding levers for facilitating lasting organizational change. Forces for change are weighted more heavily at the top, starting with the environment. The environment supplies the most potent forces for change, particularly generative change. Although this may seem paradoxical, that reaction to environmental forces may result in nonreactive, generative learning; the key lies in the relationship between the environment and the transformational factors. Tichy and Devanna’s (1986) notion of "planful opportunism" is a good way to think about this interaction. After environmental forces, leadership, mission and strategy, and culture are the most important factors for transformational organizational change.

Although the Burke-Litwin model is a useful causal model for diagnostic purposes, it does not elaborate on the specific means by which change occurs. That is, the model suggests which variables to change in order to spur transformational or transactional change, but it does not elaborate on exactly how to do so. Thus, we understand the whats and the whys, but not necessarily the hows. So that we may understand the specific hows of facilitating successful organizational development and learning, the next section turns to a related stream of literature on learning organizations.

Sections

ORGANIZATIONAL LEARNING

Conceptual foundations for organizational learning were laid by Harvard’s Chris Argyris and MIT’s Donald Schön (Argyris, Putnam, & Smith, 1985; Argyris & Schön, 1974, 1978, 1985; Schön, 1983). Argyris and Schön have written extensively on learning through action and reflection, challenging assumptions and implicit theories, and working on the processes by which problems are defined and solved. Following is a brief summary of their work, then a review of a more recent approach developed by Peter M. Senge (whose work was also influenced by Argyris and Schön).

A previous paper (De Smet, 1997a), drawing primarily from Argyris and Schön, used the adaptive systems framework, presented above, to outline the essential steps of organizational learning: (a) goal setting, (b) model building, (c) hypothesis testing, and (d) learning and action.

Goal setting is the first step to learning. In different ways, it drives both transformational and transactional change. Transactional improvement is by nature incremental and entails solving a series of operational problems over time. This is what Argyris and Schön (1978) call "single-loop" learning. Double-loop learning, on the other hand, is discontinuous. Instead of searching for new action strategies for achieving a given goal, double-loop learning occurs when the strategic or "governing variables" (Argyris et al., 1985) that frame the operational problems are themselves revisited. Schön (1983) calls this "problem setting" and explains the importance of this process as follows:

When we set the problem, we select the "things" of the situation, we set the boundaries of our attention to it, and we impose upon it a coherence which allows us to say what is wrong and what directions in the situation need to be changed. Problem setting is a process in which, interactively, we name the things to which we will attend and frame the context in which we will attend them. (p. 40)

Thus, the first step in examining a particular problem is to frame it, at least in a general sense, by setting the goals of the situation.

The second step is to build causal models based on the implicit theories of the leaders, managers, and decision makers who run the organization. For some time, cognitive psychologists have understood that humans carry around in their minds a set of vocabularies and categories, scripts and schemata, and other implicit forms of knowing and understanding the world around them (for a more complete discussion, see Nisbett and Ross [1980]). This is what

Schön (1983) calls "tacit knowing" (see also Argyris et al., 1985; Kuhn, 1962). The model building step aims to make tacit knowledge explicit so that it can be systematically explored and tested.

Step 3: Hypothesis Testing

Once identified, naïve theories and mental models are put to the test. Testing hypotheses requires not only measuring operationalized variables, but also making changes and experimenting so that alternative explanations can be explored and eliminated. Although strict scientific experimentation is not necessary, some degree of systematic inquiry is desirable. Even lacking scientific rigor, the process of systematically acting and reflecting is likely to generate feedback and learning (Argyris et al., 1985; Argyris & Schön, 1978; Schön, 1983; Weick, 1979). It is important to remember that the goal here is negative feedback. That is, we aim to disconfirm some of our hypotheses, or at least to use feedback to extend or generate new hypotheses. If hypotheses are completely confirmed and the data cannot be used to revise them, the process has failed to produce information that can be used to improve performance.

Finally, hypothesis testing is expected to result in improved organizational functioning. Improvement of functioning, and therefore learning in the organizational sense and not just in terms of individual knowledge, requires action. Without action to improve functioning, learning cannot be said to have taken place.

There are three ways that feedback can lead to improvement. First, based on level 3 feedback (complex self-regulation), incremental improvement can be achieved through better activity coordination and resource allocation decisions. The same things are done, but they are done in incrementally different amounts, sequences, or combinations.

The second way to improve functioning is based on the level 4 model (self-improvement), which can be thought of as operational problem solving or single-loop learning. Here, failure to achieve intended results through specified means results in a search for better means, but the problem to be solved is seen in exactly the same way. Improving existing means or creating new means to achieve the desired ends as originally framed is Argyris and Schön’s (1985) definition of single-loop learning.

The third way to improve, based on level 5 functioning (adaptation), is through transformational, discontinuous, adaptive change. Here, unexpected results contained in the feedback lead to a new understanding of the organization’s purpose or a fundamental reframing of the problem under consideration. This is essentially how double-loop learning is defined.

Although a general description of how organizations learn provides a useful starting point for understanding organizational learning dynamics, it does not offer a great deal of insight about how to achieve systematic learning. What must an organization do, specifically, to institutionalize the learning process? In his landmark book, The Fifth Discipline: The Art and Practice of the Learning Organization (1990), Peter Senge answered by proposing five critical elements, or disciplines, that facilitate systematic learning, growth, and adaptation: (a) personal mastery, (b) mental models, (c) shared vision, (d) team learning, and (e) systems thinking.

By personal mastery, Senge is referring to learning at the individual level. Although individual learning does not guarantee organizational learning, it is a prerequisite. If the members of an organization do not strive to learn and develop, neither will the organization to which they belong.

Building on the idea of cognitive psychology and tacit knowing, as discussed above, Senge argues that the primary reason why good ideas are not recognized, and why they often fail even when they are recognized, is problematic mental models. Our mental models, our implicit beliefs and assumptions about the way things are, are often unconscious, but they guide our behavior nonetheless. Deeply ingrained mental models shared by a group or organization are not dissimilar from the concept of organizational culture. The way we approach a problem or task is constrained by what we assume without question to be true.

According to Senge, the key to mental modeling as a discipline lies in the reflection and inquiry skills whereby organization members learn to identify, question, and challenge their implicit assumptions and overcome the defensive routines that tend to protect these assumptions (Argyris & Schön, 1985). We must strike a balance between what Argyris calls "inquiry and advocacy." When organization members spend too much time asking questions (inquiry), unsatisfied until they know everything about everything, they may be paralyzed and fail to act. On the other hand, when they concentrate only on their own position and the current way of doing things, never stopping to question its merits (advocacy), they fail to learn and change. The key is to balance reflection, inquiry, and learning, on the one hand, with planning, advocacy, and action, on the other.

Although a personal vision and individual learning are important, learning at the organizational level requires that they overlap. The individual members of an organization must share a vision for the organization. What is the basic purpose of the organization? Where will it be, ideally, in 10, 20, or 30 years? Traditionally, these questions are answered by senior management, but Senge argues that the top-down approach tends to kill organizational learning. A good vision, generated at the top and communicated to the rest of the organization, may engender compliance, but it will not garner true commitment from the organization’s members. To get real commitment and the creative energy that comes with it means that the vision must evolve through a series of dialogues at all levels of the organization. (For a theoretical distinction between compliance and commitment, see Kelman [1958, 1974].)

Developing shared vision is not the only task. Senge argues that the "governing ideas" must be mutually created and shared throughout the organization as well. By governing ideas, Senge is talking about not only the corporate vision, but also its purpose and direction (reflected in the mission statement) and core beliefs and values (i.e., culture). Creating shared vision, purpose, and values is not an easy task for leaders of organizations that do not already have them. Senge emphasizes that, to be successful, the leader’s role must be that of designer, steward, and teacher, rather than general or captain.

Like the synergy illustration provided above, Senge’s concept of team learning focuses on the synergistic qualities created through group dialogue and discussion. Mental models and defensive routines often get in the way of constructive dialogue and discussion because our habitual ways of behaving and interacting restrain our ability to challenge assumptions and propose new ways of doing things (Argyris & Schön, 1985). When we are prisoners of our own assumptions and defensive routines, current reality tends to be viewed as the only possible reality, and learning becomes impossible.

Another of Senge’s points on team learning is that it is a team skill that must be used and practiced. A group of talented individual learners who are motivated to learn is not enough for learning to occur at the group level. As with the synergy example, learning groups must learn how to learn together.

Much of Senge’s discussion of systems thinking already has been addressed. For example, Senge’s "second law" of the fifth discipline states, "The harder you push, the harder the system pushes back" (1990, p. 58). This is similar to Lewin’s notion of driving and restraining forces. Senge, like Lewin, advises that adding driving forces tends not to be very helpful, and that the key to change lies in reducing what he calls balancing forces (what Lewin called restraining forces).

Senge also points out that "today’s problems come from yesterday’s solutions," "cause and effect are not closely related in time and space," and "the cure can be worse than the disease." Senge is emphasizing two key points. First, short-term fixes are often even bigger problems later on, highlighting the importance of examining both short- and long-range organizational outcomes. Second, managing complex organizations is not about finding simple solutions to simple problems. Our assumption that a root problem is not far away from its symptoms often misleads us.

A third important point is that "small changes can produce big results—but the areas of highest leverage are often the least obvious" (Senge, 1990, p. 63). Like the levers in the Burke-Litwin model, leverage is not always easy to find, but if it is found, lasting and substantial change can result.

Sections

CONCLUSIONS

The adaptive systems framework can be used to synthesize and integrate current literature on organizational change and organizational learning. Although clearly related, recent literature on organizational learning (e.g., Senge, 1990) has remained largely separate from parallel work on organizational development and transformational change (e.g., Burke, 1994). The similar yet fragmented work in these areas points out the need for a parsimonious, theory-based, and precisely defined conceptual framework to integrate the two approaches. The adaptive systems framework can serve this purpose.

Recall that the five levels of the adaptive systems framework are not discrete levels of functioning. As shown in Figure 9, these levels can be thought of as defining points on a continuum of functioning. Transactional change occurs between level 2 (simple self-regulation) and level 4 (reactive learning), whereas transformational change occurs between level 4 (reactive learning) and level 5 (generative learning). The five levels of functioning also delineate four different learning dynamics: (a) continuing operations, (b) resource management, (c) innovation, and (d) adaptation (for a thorough description of the adaptive systems framework continuum depicted in Figure 9, see De Smet [1997b]).

The integration of work on organizational change with organizational learning using the adaptive systems framework provides a theoretical basis on which to make a series of predictions about how and when particular organizational dynamics will lead to effective organizational management, innovation, and adaptation, and when each of these sorts of changes will be successful given key internal and environmental conditions. The following section, applies and extends these predictions as they relate to substance abuse treatment organizations.

Sections