------ # Understanding Command and Control **Written By:** - David S. Alberts **See:** Alberts_UC2.pdf --------- ## Nature of Understanding There are many different ways to explain the concept of understanding, each with its own nuances. To first order, to understand something is to be able to grasp its nature or significance; to understand is to comprehend (an idea or a situation); to understand is the ability to offer an explanation of the causes of an observable state or behaviour. In our past work, we have stressed that “understanding” goes beyond knowing what exists and what is happening to include perceptions of cause and effect, as well as temporal dynamics. Since the dawn of empiricism, understanding has been associated with systematic observation, experience, and expertise rather than revelation. We say that we understand something when the result seems reasonable to us and we say that we do not understand it when the result is unexpected or (at least to us) without a logical explanation. Understanding resides in the cognitive domain and, like everything in the minds of humans, is subjective, influenced by perceptual filters and biases. However, one’s understanding may not be “correct,” that is, it may not conform to objective reality. Thus, one can apply attributes to understanding that correspond to the attributes we associate with information, including correctness and completeness. To understand something does not mean that one can predict a behavior or an event. Prediction requires more than understanding, thus even if one understands a phenomenon, one may not be able to predict, with anything that approaches a level of usefulness, the effect(s) of that phenomenon. Prediction requires actionable knowledge, specifically the values of the variables that determine (or influence) the outcome in question. Operationally, the most that can be expected is to identify meaningfully different alternative futures and indicators that those alternatives are becoming more or less likely over time. Understanding is also insufficient to improve a situation. Improvement that is deliberate and not the result of trial and error requires both the ability to predict and the ability to control the values of some or all of the variables that affect the outcome. Thus, the value or utility of understanding in order to improve a situation depends upon specific knowledge and the degree to which one can control or influence key variables. ## Degrees of Understanding There are degrees of understanding that correspond to a scale that runs from a cursory understanding to a complete understanding. In terms of understanding Command and Control, a cursory understanding of C2 would involve understanding only what C2 is trying to accomplish, that is, the result that C2 is designed to achieve. A greater degree of understanding requires recognition of the different C2 Approaches and their applicability. The degree to which one can answer the following questions about C2 corresponds to the degree to which one understands its nature and its application to selected situations. - What are the possible Command and Control Approaches? (how desired results could be accomplished) - What are the key differences among Command and Control Approaches? (the dimensionality of the C2 space) - What influences the ability of a C2 Approach to realise its objectives? - Which approaches are appropriate for a given set of circumstances? - What can be expected if a particular approach is adopted and a specific set of circumstances is obtained? Despite the fact that military organisations have practiced Command and Control for millennia, the answer to even the first of these (i.e., possible approaches) is not definitively known because military organisations have, until very recently, only explored a small subset of the approaches that appear to have potential. This book seeks to provide a conceptual foundation that can be used to develop a better understanding of Command and Control so that answers to these questions can be found. One of the biggest problems is that there has been relatively little effort expended on finding answers to some of these questions because of a prevailing view that we have a C2 Approach that works well (or that it is thought to have served us well so far). In fact, the view that traditional Command and Control Approaches have worked well is debatable and the view that traditional approaches will continue to serve us well is not supported by current events and operations. The relevant threats, operating environments, technologies available, and our understanding of human enterprises are all changing. ## Facts, Theories and Models If we are to improve our understanding of Command and Control, then we will need to establish facts, develop testable theories, and instantiate these theories in models. In short, we must build a body of knowledge, gain experience, and develop expertise. To accomplish this, we need to observe reality, intellectually develop conceptual models, and design and conduct experiments to calibrate and validate these models. This entails the collection of empirical evidence, the conduct of analyses, the publication of results, and the archiving of data. These tasks are iterative. A complete set of facts is not necessary to formulate a theory or construct a model. Theories and models are most often a mix of what we know (or think we know) and what we think (conjecture or hypothesise). Theories are almost always conceived from a limited understanding (having only a fraction of the necessary facts and the relationships among them) and serve to focus our efforts to identify additional relevant variables and to discover relationships. A fact is a piece of information having objective reality, and facts reside in the information domain, but how individuals and groups interpret facts is another thing. These interpretations or perceptions occur and reside in the cognitive domain. Therefore, theories that address human behaviour must deal with both facts and the ways in which they are perceived. A theory consists of the abstract principles of a body of facts. The dictionary notes that the term theory can be applied to both a science and an art (as in music theory). Given that many think of Command and Control as an art and a science, the notion of a “theory of C2” would be appropriate in either case. A theory and a model that instantiates a theory consist of a set of facts (or assumptions) and the relationships among them. A theory or model can be as simple as the economic price theory we learned in Economics 101: \[P = f (SD)\], where P = price, S = supply, and D = demand. However, operationalising this theory is far from simple and has occupied many economists for a long time. Theories and the models that instantiate them are representations for a purpose. We were first introduced to models as children. Dolls, toy cars, guns, and swords are iconic models (physical representations of the real thing). Iconic models are also used extensively to test designs for ships and airplanes in tanks and wind tunnels. This allows us, at relatively low cost, to subject these designs to various conditions and observe their behaviours. It allows us to go beyond what we normally experience in the real world and test these designs under extreme conditions. Additionally, iconic models need not be complete representations of reality; they only need to provide an adequate representation of the characteristics that we are interested in for the purpose of the experiment. In the tank or wind tunnel, this may be only the shape of a hull or fuselage. Iconic models are relatively easy to build and they are easy to relate to the theory or object they are designed to represent, but they are not easy to change. If one wants to explore a series of hull shapes that represent changes to a particular parameter, then many models need to be built, each one representing a different value of the parameter in question. Testing the effect of the value of the parameter in question thus involves running a series of tests on each model. One can see how working only with physical models might be a very time consuming process and hence limit exploration of a parameter. Exploring multiple interrelated parameters would be even more cumbersome. Different kinds of models are better suited for simultaneous exploration of the effects of a number of parameters. These more agile models are mathematical models and simulation models, both of which are instantiations of a conceptual model. These models are designed to allow for the changing of the values of a parameter or a set of parameters and then determining the effect that this has on the variables of interest. Conceptual models are representations of how we think (conceive) about something, in this case Command and Control. The building blocks of these models are concepts, which translate into one or more variables and the relationships among them. The degree of specificity with which these relationships are expressed in conceptual models varies from the existence of a relationship or influence to a more definitive expression of the nature of the relationship. Conceptual models are often depicted graphically with the concepts expressed as boxes or other shapes and the relationships between and among the concepts as lines or directional arrows. Mathematical models consist of sets of related equations. Conceptual models, mathematical models, and simulation models all have the same basic building blocks: variables and the relationships among them. Conceptual models and mathematical models are not working expressions, while simulation models are, in reality, tools that bring conceptual models or mathematical models to life, producing outputs from a set of inputs. Other tools serve this purpose as well, including dynamic and linear programming, expert systems, and the familiar but versatile spreadsheet. A major difference among these types of tools is whether they are event-oriented, rule-oriented, or instantiating formulas. Simulation models sometimes do all three, generating events on a predetermined or stochastic basis, having agents that employ rules that govern decisions, and calculating the values of parameters using static or dynamic formulas. ## Building a Conceptual Model Despite the ease of constructing a conceptual model, that is, going to a whiteboard and drawing a bunch of shapes and lines, building a meaningful conceptual model is quite difficult. The most important decisions involve what to include and what not to include. When a piece of Mozart’s was criticised for having “too many notes,” the composer replied that the piece did not have too many or too few notes but exactly the right number of notes. So too does a model that is “fit for use.” The important thing to consider is whether or not the model serves its intended purpose. Well-conceived and constructed models do not have too many or too few variables, but just the right ones. The number of variables must be kept to the bare minimum needed in order to enable the model to communicate its concepts to others. For this reason, less is more. Keeping the number of variables and relationships under control makes the model as simple as possible and thus as easy to understand as possible. This requires the model to extract the essence of reality, and only the essence. The way in which designers of a conceptual model balance the need for simplicity with the need for fidelity often determines success. One way of dealing with these conflicting objectives is to have a number of depictions or views of the conceptual model, each of which serves a specific purpose and a way of organising detail. ## Identifying the Minimum Essential Concepts To illustrate this point and the nature of a conceptual model, we have built a relatively simple model designed to explore the control of a room’s temperature. Figure 1 depicts a set of the minimum essential concepts (one or more variables that represent the necessary elements of the model) needed to explore the approach taken in an attempt to keep the temperature of a room within desired bounds. ![[UC2-Fig-1.png]] The minimum essential elements include: command, control, two sets of capabilities (heating and cooling), the target (room), its environment, and a sensor (thermometer). In its most common instantiations, command involves the determination of the desired temperature and the function of control is actually built into a rather simple thermostat. The function of control translates the desired temperature into a set of rules that govern what actions are taken. As we will later define these terms, the thermostat embodies elements of both control and sensemaking. If the temperature behavior in the room does not meet expectations, a number of actions can be taken. Command may decide to reset the desired temperature, buy a new or different thermostat (hence changing the nature of control), modify, repair, or replace one or both of the systems (hence, for example, reallocating resources), just to name a few of the possibilities that create a different set of conditions. This simple model can represent a wide variety of Command and Control Approaches, help us to understand the nature of the task involved, and inform a wide range of decisions. ## Instantiation of concepts in a mathematical model How much we need to know about any one of the model’s conceptual elements depends on (1) the nature of the purpose or use of the conceptual model and (2) reality. For example, if the environment were invariant, then the effect of the environment on the room temperature could be accounted for as part of the characteristics of the room and for all intents and purposes would not need to be depicted separately. If the characteristics of the room were also invariant over time, then all we would need to represent the concept of the room would be two functions: one for heat loss over time, the other for heat gain over time, both of which would be conditional on the current room temperature. If the sensor reported room temperature accurately, instantaneously, and with a precision that was appropriate, then we would not need to represent the sensor. In a similar fashion, the heating and cooling systems can be represented by mathematical expressions that are functions of current temperature and time since the heater/air conditioner was turned on. Combining these mathematical expressions for the systems and the room characteristics into a single expression is rather straightforward. Thus, room temperature at time t may be expressed as T(t) = T(t- t) + (t- t), where , the gain/loss of room temperature between t- t and t, is a function of the ability of the heater/air conditioner to warm/cool the room at time t- t, the room temperature at time t- t, and both room and environmental conditions. We now turn our attention to the nature of the Command and Control Approach that is being considered. First, as we earlier assumed, the function of command consists of picking a desired temperature for the room and the approach to control consists of a translation of this intent into a simple decision, namely to turn the heater on when the room temperature falls to a predetermined temperature, turn the heater off when the temperature is at or above a predetermined temperature, turn the air conditioner on when the room temperature rises to a predetermined temperature, and turn the air conditioner off when the room temperature is at or below a predetermined temperature. The function of control is to select these predetermined temperatures. Figure 2 instantiates these assumptions in the form of a mathematical model. This mathematical model is deterministic, that is, the behaviour of room temperature is totally determined by the values of the parameters embedded in the temperature gain/loss function. ![[UC2-Fig-2.png]] This very simple model can be used to investigate a number of important questions. Given the assumptions: the room and environmental characteristics, the performance characteristics of the heating/cooling systems, and the Command and Control Approach, - How much of the time can the room temperature be maintained at x degrees plus or minus y degrees? - If the characteristics of the environment were to change (specify the change), what effect would it have on the ability to maintain a given room temperature? The first of these questions involves merely plugging in the value of the target temperature and calculating room temperature as a function of time. The second question involves a revision to the formula, T(t) = T(t- t) + (t- t), in effect, a modification to the mathematical model. Note that no revisions to the conceptual model are required. The way in which command has been represented in this example equates to command by intent, a statement of the desired outcome. The approach to control is interventionist,16 that is, specific orders are given to the “forces” (the heater and the air conditioner) at irregular intervals when the temperature reaches x (a specified event or condition). Thus, control here consists of a simple decision linked to scripted behavior. These types of decisions can be easily automated (with a thermostat, they are built in as hardware, or with a more sophisticated thermostat, a combination of hardware and software). ## One Conceptual Model, Many Instantiations While the above instantiation of the conceptual model is suitable for the purposes defined above (providing answers to the questions), it may not be suitable for addressing a different set of questions. However, as long as the conceptual model is suitable, the mathematical model may be altered to reflect changes in assumptions. For example, the nature of command intent could be altered from a simple target temperature to a target temperature that changes with time or circumstances. For example, the room could be maintained at a specific temperature during working hours and a different temperature the rest of the time. A more sophisticated behaviour may be desired, for example, a temperature target for the room when occupied and a different target for the room when empty. This would require modeling a sensor that was capable of detecting whether the room was occupied or not, but the basic structure of the original conceptual model would be the same. There are, of course, ways to alter the temperature of a room other than simply turning a heater or air conditioner on or off. For example, windows and doors could be opened and closed, shades raised or lowered, and lights turned on or off. In addition, the number of people occupying the room could alter its temperature. To this point, the model has focused only on the variables found in the physical and information domains. The physical domain is the source of the equations regarding temperature changes; the information domain consists of reports of room temperature. Now let us consider an alternative way of defining the room conditions that we seek (desired values). Instead of using a measure that can be determined by physical measurement, like room temperature, one could use the comfort level of the room’s occupants. While there is a relationship between room temperature and comfort level, comfort level will vary from person to person and with other factors including humidity, air movement, and light levels. They will also be affected by apparel choices. Will the need to consider means other than heating and cooling of the room’s air (e.g., to affect humidity or light levels) affect the way we need to formulate our conceptual model? If we look at the conceptual model, we see that the means identified to change room temperature includes a heating system and a cooling system. Generally, we would interpret a heating system to consist of an oil, gas, or electric furnace with air ducts, or perhaps a radiant system with steam or hot water or a heat pump. There are of course other possibilities that might leap to mind such as an active or passive solar heating system. If we liberally interpret the terms, a heating and cooling system may be powered in any of a variety of ways and also include the ability to change airflow, light conditions, humidity, and the degree to which the room “membrane” isolates it from the environment. If we broadly interpret these terms, the conceptual model, as we have formulated it, remains appropriate for our purposes as far as the way we can alter the physical characteristics of the inside of the room. Will the adoption of “comfort level” as the measure of desired room conditions require us to change the conceptual model? The conceptual model includes the existence of a sensor that reports on the condition of the room. In our initial discussion of the problem, we interpreted this to mean a thermometer. If there was such a thing as a “comfort sensor,” we could simply replace the temperature sensor with a comfort sensor, but of course there is no such generally accepted device. At this point, we have three options. The first is to include a set of sensors that measure a variety of physical factors that are known to affect an individual’s comfort level to obtain an indicant of comfort. The second is to indirectly measure the comfort level of individuals occupying the room by their behaviour (e.g., sweating or teeth chattering). The third is to have the occupants directly report their level of comfort. In each of these cases, all we have done is to specify what we mean by “sensor” and thus the conceptual model, as formulated, remains appropriate. How will the adoption of “comfort level” change the way we look at Command and Control? If we use a set of measurable physical characteristics as an indicator of comfort, we need only (1) change the way command intent is expressed (i.e., target values for a defined function of the set of variables), (2) adjust the simple decision accordingly, and (3) map it to a new set of actions related to the means of affecting the characteristics of the inside of the room (e.g., lowering a shade). Again, this does not require a change in the way we have formulated the conceptual model. Dealing with each of these changes to the way that we think about how we affect the conditions inside the room and how we value the outcome we have achieved does not require a change to our conceptual model. However, they do require us to identify new variables and relationships and the introduction of the value metric of “comfort” requires considerations that involve the cognitive domain. Let us now consider a more radical way of controlling room comfort. Suppose instead of trying to sense the comfort level of individuals occupying the room and taking appropriate action, we provide them with the means to do it themselves. With access to the appropriate means, occupants can adjust the heating and cooling systems, adjust air movement using fans or room openings, change humidity levels, and adjust lighting conditions. Adopting this approach clearly decentralises control. The decentralisation of control can lead to actions being taken that may conflict with one another. Therefore, we need to deal with the interesting question of how two or more occupants could accomplish the functions associated with command. However they decide to do it (delegate, negotiate, collaborate), the interactions between and among the occupants would be in the social domain. Once again there is no need to revise the conceptual model, but once again new variables and relationships need to be introduced. We have come a considerable way from our original notion of a room’s temperature being controlled by what amounts to a thermostat set to a given level. Some of the complexity that we added (new variables and relationships) was a result of a change in the intended use of the model, some by a recognition that the assumptions were not appropriate, some by changing the way we define the room conditions we seek, and some by altering our Command and Control Approach. ## Process Versus Value Views At this point, the model formulation addresses all four domains: physical, information, cognitive, and social. It can be used to describe the conditions in the room over time and compare these room conditions to a specific target condition or set of conditions. As formulated, this model has only one embedded measure that we can use to characterize success: the nature of the difference between a desired outcome and the actual outcome. The reason that the model contains this measure is related to process rather than value. This comparison is required for the processes of Command and Control. The target conditions are an expression of the output of the process we call command and an input to the process we call control. The other input to control in the model is a reading of room conditions. This model, as formulated, can be used to explore different heating and cooling methods and to see what their effects are on the ability to keep room conditions in sync with a dynamic target. This model can also be used to determine what range of environmental conditions can be tolerated Suppose that we wanted to know: - How much energy does it take to keep the temperature within specified bounds? or - Does it make more sense to better insulate the room or increase the capacity (effectiveness) of the heating/cooling system? The first question requires some additional information about how much of the time the heater is on and how much of the time the air conditioner is on, and how these relate to a measure of energy consumption. Energy consumption can, in turn, be related to cost. The second question requires an understanding of the costs and system performance improvements associated with each of the two ways of improving our ability to maintain room conditions. In general, these issues assume that there are things we can control and that our choices make a difference in our ability to create the effects or achieve the goals specified by command intent. Beyond our ability simply to realize intent, there are issues of efficiency and/or cost associated with our choices. There is an implied value chain that links the quality of command intent and its expression to system performance and on to measures of value associated with outcomes. Information quality clearly affects each of these. Making the value chain explicit helps us to focus on what really matters rather than on describing behaviors. Figure 3 identifies the value metrics associated with each of the concepts in the conceptual model. ![[UC2-Fig-3.png]] Value metrics, like concepts, can consist of one or more variables. For example, the quality of the information provided by a temperature sensor is determined by, at a minimum, its accuracy, currency, and precision. A meat thermometer that provides you with an instantaneous readout that correctly tells you that the meat is well-done, medium, or rare may or may not be as useful as one that tells you, within one degree of accuracy, that the meat was 140 degrees 30 seconds ago. The issue at hand is “fitness for use” and that in turn may depend on the nature of the situation (e.g., rate of temperature change, or the degree of education and experience of the cook). The inclusion of value concepts, in addition to process concepts, opens the door to a richer set of uses for the model. As the above example shows, if we get the basic concepts right, we will be able to develop a series of instantiations that helps us to deal with a variety of issues as well as to incorporate both facts and the relationships among them as we gather empirical evidence and understand its implications. Instantiations of the conceptual model, in the form of specific mathematical or simulation models, help us focus on what is important for the problem(s) we are working on. The above example is meant to acquaint the reader with the basics of model building that could be used to explore a set of issues rather than modelling that is so specific that one needs to start from scratch if an important aspect of the problem or our understanding of the problem changes. This is what happens when one skips the development of a conceptual model and goes directly to a specific instantiation. All too often, simulation models have built-in (hard-wired) sets of assumptions or represent a partial formulation of the real problem (a limited set of concepts or variables). Conceptual models represent our current state of understanding and provide a firm foundation to test and improve our understanding. Without a conceptual model to serve as a means to organise what we know, efforts to improve our understanding will be less efficient and less effective. The next chapter deals with the nature of Command and Control. With the understanding of what a conceptual model is and why we want one from this chapter, and a basic understanding of Command and Control from the following chapter, the reader will be well-prepared for the step-by-step development of a conceptual model of Command and Control that begins in Chapter 5.