As introduced in Chapter 2, this Appendix provides additional details of the committee’s ecosystem model conceptualization of the modern U.S. Air Force (USAF) human capital management (HCM) system (shown in Figure 2-2) and externalities that impact that system. To create the ecosystem model, the committee used a causal loops diagram to model the USAF HCM ecosystem and illustrate the causal links between the system’s various components as is described in more detail here.
SYSTEM DYNAMICS MODELING
For very large and complex systems, developing an understanding of how disparate parts of the system interact with each other can be useful. A category of modeling that comes from the operations research community and has grown in popularity over the past 50 years is known as System Dynamics (SD) modeling. There are two principal components of SD modeling: causal loop diagrams (CLD) and stocks and flows. Both components bring a way of exploring the dynamics of behavior in a system. As noted by Jay Forrester (1969, quoted in Sterman, 2000):
Like all systems, the complex system is an interlocking structure of feedback loops . . . This loop structure surrounds all decisions public or private, conscious or unconscious. The processes of man and nature, of psychology and physics, or medicine and engineering all fall within this structure (Sterman, 2000, p. 107).
The structure of a system is a set of connections between components. The aim of SD modeling is to capture the dynamics of this system by mapping these connections. The effect of one component on another can be deceptively simple: increasing salaries attracts more applicants; decreasing safety controls increases injuries. But when an enterprise full of these simple interactions is joined together in a single model, the resultant complexity becomes clear. With such a model, one can create simulations that capture the mathematical nature of the interactions, the delays inherent in the effects, and the ability to randomize external inputs. This provides a powerful way of grasping the subtleties of enterprise-wide dynamics, and thus aids decision-making.
Creating a model that captures the dynamics of a very large enterprise is a time-consuming effort. For this study, the committee used a CLD in order to quickly capture the observed dynamics in a consumable form. Follow-on work to validate the relationships and develop the underlying mathematics of the relationships is left to future researchers. The current model’s value lies in its illustration of the complexity of the system, including important feedback loops and the extensive impact of any single decision. It shows how a simple change in one variable can have widespread “downstream” consequences that a complex enterprise would want to understand and take into account when considering such a change.
CAUSAL LOOP DIAGRAMS BASICS
A causal loop consists of multiple causal links that show the effects of change on elements in the system. A causal link is represented with four graphical components: two variables, an arrow, and a polarity mark. The arrow originates at one of the variables and ends at the other variable, with the arrowhead indicating the direction of the relationship. The polarity mark, normally denoted by either a plus or a minus symbol, indicates the general nature of the relationship.
The fundamental underpinning of the variable relationship is that the two variables form a dependent relationship within a broader system. The value associated with the dependent variable is affected by changes in the first variable. For example, consider the classic1 example of birthrates and population: when birthrates increase, population increases as a result, more than it normally would if the rate had not changed. The relationship indicated by the polarity mark, a plus symbol, is positive: the direction of change is in the same direction. The change could be increasing or decreas-
1 This example is widely used in the literature, simply because it is so easy to understand. A more in-depth treatment can be found in J.D. Sterman, Business dynamics: Systems thinking and modeling for a complex world (2000).
ing: when birthrates decline, the population declines. The point is that the change is in the same direction in the dependent variable as in the first variable. Figure B-1 shows this positive causal link.
A classic example of a negative relationship, one in which the effect is in the opposite direction, is that of death rate and population. As death rate increases, population decreases faster than it normally would. Alternatively, as death rate decreases, population increases from the expected value, had death rate not changed. Figure B-2 shows that relationship.
These two links can be combined to create a simple causal loop, illustrating how complex system behaviors can emerge from elemental analyses. Both birthrate and death rate affect the population levels, and create two feedback loops, shown in Figure B-3. The growth rate on the left is held
in check by the baselining effect of the death rate on the right, but only to a limited extent. This is an overly simplistic view of this situation, of course. Other important complications include food supply, warfare, medical treatments, and more. The purpose here is to introduce the concept of feedback loops.
One of the reasons this example is used so frequently is because it illustrates an important point of this modeling method: the values exist for variables before the effect of change is considered. There are births, there are deaths, and there is a population. If the birthrate were to increase or decrease, what would be the resulting changes in population and death rates? If the population were to increase, what would be the resulting changes in birthrates and death rates?2 This is the critical point of this method: it allows the modeler to simulate changes and discover the cascading effects throughout the modeled system.
The more complicated a system, the more beneficial an SD analysis can be. The benefits include identifying potential problems, but extend to showing how vastly different parts of the system can exert influence in cascading effects. Managers can use the analysis as a basis for organizational, process, and technology strategies, as well as for designing checks and balances, reducing conflict, and increasing effectiveness throughout the system.
THE USAF HUMAN CAPITAL ECOSYSTEM MODEL
As introduced in Chapter 2 (see Figure 2-2), the committee’s hypothesized USAF human capital ecosystem model illustrates the challenge of managing the various portions of the system, all of which demonstrate that similar task groupings have very different needs in terms of skill sets, focus areas, and structure. And yet each part of the system affects every other part of the system, sometimes directly but mostly indirectly.
The structure shown is a flattened version of the overall model, which actually more closely resembles a sphere. In order to flatten out the ecosystem model, “shadow variables” are used in several places. The shadow variables are visually different than the regular variables for ease of recognition (grey italics). There are several variables that are shadowed on the diagram in order to reduce visual complexity, including mission effectiveness and needs of the Air Force (AF). The shadow variable is mathematically and structurally exactly the same as the original variable. By using the shadow of the original variable, the same variable can be linked to multiple
2 Birthrates and death rates are defined per year, rather than per capita. For example, if death rate for a population of 100,000 is 100 people per year, then the fractional death rate is 0.001. If the population increases (through more births or immigration) to 200,000 and the fractional death rate remains the same, the deaths per year will increase to 200 per year.
other variables without links that cross other links. The use of the shadow variables allows complex links to be created without making the diagram too messy to read. For example, the needs of the AF are central to the entire diagram but the variable is also shadowed in the bottom right-hand corner of the flattened diagram. By using the shadow, it is possible to create links that do not overly intrude on other parts of the model.
The color-coding is designed to enhance readability as well. The colors used are as follows:
- Yellow: used to highlight the important issues, such as the needs of the AF;
- Green: used to point out where research is done to improve the overall effectiveness of functions;
- Blue: the goal of excellent human management efforts—measures of effectiveness;
- Pink and peach: the people-focused externalities—applicants and economic conditions; and
- Red: the externalities with the potential to force important reactive change.
There is no starting point to this model: everything is related to something, directly and indirectly. But one must start somewhere in trying to understand the relationships. For the purposes of this report, one useful way of looking at this model is by starting with the category mission effectiveness. This is, after all, the fundamental concern of HCM. How does the Air Force create a force structure that is competent and capable of executing missions and defeating our adversaries? Figure B-4 zooms in on this part of the diagram, putting mission effectiveness at the center. (Note that the larger ecosystem model also includes mission effectiveness as a shadow variable in other physical parts of the model [e.g., top center].)
Mission effectiveness measures the operational success of people and equipment to execute the mission. The effectiveness of people is directly related to their level of competency to execute a particular mission and the effectiveness of the team working together to execute the mission. When the aggregate level of competency of the service members increases, it follows that the mission should be more effectively executed. However, it is also clear that team dynamics plays a very important role in team operations: the more effectively a team can work together, the more likely its members will be effective in executing a mission.3
3 As further explanation, as training availability increases, team effectiveness should increase assuming that training includes working as a team. The same applies to Aggregate Competency Level: the more training, the higher the level of competency. Some jobs are performed indepen-
An excellent level of competency for each service member does not appear magically: it comes from training and from putting the service member into an appropriate career field. Figures B-5 and B-6, discussed below, show how these variables affect competency.
Using Figure B-5, several hypothetical situations can be extrapolated that illustrate the impact of training effectiveness and training availability on aggregate competency level. For example, if training is not effective, service members will not achieve the level of competency that they might have with better training.4 If training is not available or is only periodically offered, then the level of competency will suffer. If either training effectiveness or the level of competency of service members decreases, then more research into training outcomes is needed, which should improve training
dently with no team required. Some jobs are team oriented, strongly or weakly. A team that is fighting among itself will not perform as well as a team that gets along, everyone understands how to operate together, and nothing needs to be said in order to get the job done correctly.
4 This issue came up several times during the committee’s data-gathering site visits, specifically in anecdotes regarding maintenance training whereby Airmen were trained in a particular aircraft model series only to be assigned to work on a different series with sufficient differences to render the training largely insufficient (or another aircraft model entirely).
effectiveness. If externalities such as the rate of technology change or operational disruptions occur, training will need to change in order to keep training content appropriate to the needs of the mission. This in turn should inspire more research into how such training should be provided.
Figure B-6 explores the impact of person-job fit on the level of competency of service members.
If a person is well-suited to a job, a team, or a role, by aptitude, personality, or skill, that person is more likely to be satisfied with the job, less likely to fail during training, and more likely to be competent in the execution of the mission requirements. When the competency of the service
member is high, there should be fewer retraining needs, except as required by technology change or operational disruption. When individuals are more suited to their roles, there should be fewer cases where a service member is removed from a role for poor performance, which should result in fewer unfilled job positions. It is important to note that there will always be a certain number of unfilled job positions; the effect considered here is how change increases or decreases that level.
Two research areas inform the person-job fit category: assignment selection research and attrition cause research. If people who are otherwise well-suited to a job are failing to perform up to standard for one reason or another, research into why that is occurring can lead to changes that reduce that problem. For example, if research indicates that giving people more time to develop the required bone density to avoid injury on the job reduces the attrition rate of people well suited to a career field, then changes can be considered as to how or whether to accomplish that.
Figure B-7 centers on person-job fit: While an Airman’s fit to a particular job category influences the aggregate competency level of service members and the person’s satisfaction with the job, a good fit should also reduce attrition due to preferences and aptitude to successfully complete training. In turn, person-job fit is affected by the number of unfilled job positions and the needs of the Air Force. Simply put, if there are more needs than there are people, there is a decrease in the ability to ensure everyone is in a job
that is a good match. An ensuing decrease in rates of good person-job fit would, of course, have a cascading impact on aggregate level of competency and, ultimately, mission effectiveness. But in the end, filling job positions must be prioritized over ensuring a good person-job fit, and the needs of the service strongly influence that calculation.
The Air Force is a very large organization and operates on long time frames. There are significant time lags between predicting the needs of the Air Force and executing a recruitment, assessment, selection, and training program. Although not specifically accounted for in the model, time lags and rates of change across the entire ecosystem would be an important consideration if the Air Force is to use the model operationally.
Figure B-8 centers on the needs of the AF, an estimate of future needs (in quantity and capability) as defined by Air Force Specialty Code, AFSC that are impacted by changes in vacant positions and the accuracy of future needs forecast. The accuracy of future needs forecast is affected by shifts in the rate of technology change and the level of research into future needs. Mission effectiveness changes should inspire changes in the type and focus of research into future needs as well.
Changes in the needs of the Air Force cause changes in recruiting efforts and how selective the service can be in accepting applicants. When the needs of the Air Force increase, the ability to be picky about who is selected
decreases. When the needs of the Air Force decrease, the recruitment and acceptance processes can be more selective.
Figure B-9 focuses on how the flow of applicants into the service is affected by several influences. Changes in demographics obviously affect the number of people who are eligible to serve, as do economic conditions. Other influences that can change the number of people interested in applying to the service include knowledge about opportunities and the reputation of life in the service. Effective recruiting efforts can increase the spread of this type of knowledge.
FEEDBACK LOOPS IN THE MODEL
The ecosystem model is filled with feedback loops, some direct and some indirect. A loop is created when behavior follows a series of links back to the original variable. The loops are characterized as having either a positive (reinforcing) or negative (baselining) effect, depending on how the changes cascade. Furthermore, each feedback loop interacts with others, making change analysis complicated. For example, the variable person-job fit category is included in 147 different feedback loops, some as short as two variables and others as long as 19 variables. Clearly, that variable is central to the overall behavior of the system. Similarly, mission effectiveness is included in 141 different feedback loops. To truly understand the dynamics of the overall model would require significant study and analysis.
Two examples of feedback loops are presented below for illustration. The first example is of a loop resulting in baselining behavior; the second shows reinforcing behavior.
Example 1: Mission Effectiveness (baselining effect)
- Research into Future Needs
- Accuracy of Future Needs Forecast
- Vacant Positions
- Person-Job Fit
- Aggregate Competency Level of Service Members
The following relationships describe the behavior of this loop. Overlapping effects as the loop repeats are in bold. All other things being equal,
- a decrease in mission effectiveness should increase the need for research into future needs;
- an increase in research into future needs should increase the accuracy of future needs forecast;
- an increase in the accuracy of future needs forecast should reduce the number of unfilled job positions;
- a decrease in the number of unfilled job positions should increase the ability for the Air Force to fit people into appropriate job categories;
- an increase in fit to job category should increase the level of competency of service members; and
- an increase in the level of competency of service members should increase mission effectiveness, which loops back to the beginning of the behavior cascade but in a different direction, causing the baselining effect.
In this loop, a tension is illustrated between mission effectiveness and managing the future needs of the Air Force. While things are going well, it is tempting to back off on future needs research, but when that occurs, erosion in mission effectiveness may well follow.
The second example illustrates a feedback loop with a positive, or reinforcing, effect.
Example 2: Person-Job Fit Category (reinforcing effect)
- Satisfaction with Job
- Job Reputation
The following relationships describe the behavior of this loop. Again, overlapping effects as the loop repeats are in bold. All other things being equal,
- a decrease in fit to job category should decrease satisfaction with the job;
- a decrease in satisfaction with the job will decrease the job reputation;
- a decrease in job reputation will decrease the number of applicants;
- a decrease in the number of applicants will decrease the ability to be highly selective;
- a decrease in selectivity will decrease the ability to fit people optimally to job categories, which loops back to the beginning of the behavior cascade in the same direction, causing the reinforcing effect.
In this loop, the reinforcing effects of prioritizing putting someone in a job that he or she is well suited for are illustrated. When it is done well, the effect cascades to allow even better fit to job category. When it is done poorly, the effect cascades in a cycle of erosion.
SUPPLEMENTAL DETAILS ON TECHNOLOGY DISRUPTION
This section elaborates on the effects of feedback loops in the ecosystem model by providing supplemental information to the fictional example contained in Chapter 2 (see Box 2-1). The additional trees given here, as well as those in Chapter 2, demonstrate the cascading effects that a single external shock can have across the entire ecosystem. These trees show the impact of a technology disruption. It is important to understand that shocks (or policy or other changes) internal to the system are capable of having a similar effect. As in Box 2-1, the items appearing twice in the tree (i.e., impact multiple points) are shown in parentheses.
Need for Training Changes
In Chapter 2, Figure 2-6 shows that the increase in Need for Training Changes causes an increase in the need for instructor retooling and new training development, but also triggers an increase in a need to do research about training outcomes. This in turn cascades to training effectiveness, which impacts variables as shown in Figure B-10.
Of these, the most significant cascading effects are seen through the link to the aggregate competency level of service members—as shown in Figure B-11—which in turn cascades to six variables, some unique, some reinforcing others.
Appropriateness of Training
As shown in Figure 2-7, changes in the training effectiveness category cascade to other variables, the most significant of which is the aggregate competency level of service members (see Figure B-11).
Accuracy of Future Needs Forecast
As shown in Figure 2-8, changes in the accuracy of future needs forecast principally affects two variables, the most significant of which is the needs of the Air Force (AF). This in turn cascades to the variables as shown in Figure B-12.
Loop Example: Accuracy of Future Needs Forecast
The effects on the accuracy of future needs forecast cascade through 137 feedback loops. One of these is described here:
- Accuracy of Future Needs Forecast
- Unfilled Job Positions
- Needs of the AF
- Person-Job Fit
- Level of Competency of Service Members
- Mission Effectiveness
- Research into Future Needs
- Accuracy of Future Needs Forecast
Graphically, these are shown in the following six trees (grouped together under Figure B-13). Each of these charts show how the changes cascade and ultimately affect the accuracy of future needs forecast.
In the modification of the ecosystem model shown in Figure B-14, this loop is shown in the white area. Note the use of a shadow variable for accuracy of future needs forecast to make the model drawing less cluttered.
These trees elaborate on the ways in which a single shock can create a cascade of effects registered across the system. The committee offers this detailed example to encourage the Air Force to conduct similar systematic considerations of the entire ecosystem to better understand the potential impact of single-point actions or decisions across the entire ecosystem.
This ecosystem model, while hypothetical and unvalidated, illustrates the power and necessity of whole system dynamics analysis. The results of such analysis can be used to strategize changes and test the effects of such changes in the entire system, paying particular attention to unintended effects. Applying validation testing techniques to this model will identify areas in which the model may be modified to more correctly reflect the reality of the system. The validation processes can also identify the limitations of the model and pave the way for future research into developing alternative views of the system, such as a mathematically rigorous stock and flow diagram.
Forrester, J.W. (1969). Urban dynamics. Portland, OR: Productivity Press.
Sterman, J.D. (2000). Business dynamics: Systems thinking and modeling for a complex world. New York, NY: McGraw-Hill Education.