Table of Contents
"When You Assume..."
Richard A. White
Orchard Avenue
PO Box 1688, Pleasant Valley, NY 12569
845-471-7681
richwhite@orchardavenue.com
This paper shows how to develop better causal loop diagrams by using simple thinking tools
borrowed from other fields of endeavor. It will be a valuable starting point for practitioners
looking to add a new dimension to their work, as well as for educators wanting to offer a richer
learning experience. That is not to say that any of the diagrams found in this paper are works of
genius, or even that they are anything more than early drafts - only that the "after" versions are
better than the "before" versions, thanks to the use of the tools described here.
I. Introduction
The Causal Loop Diagram (CLD) is a fundamental and extremely powerful tool of systems
thinking. Its accessibility makes it ideal for introducing systems thinking to the uninitiated, yet it
is robust enough that it can also be an early step in the process of developing a complex model
for simulating system dynamics.
There is, however, something that can corrupt the power of CLDs - unexamined assumptions.
The examination of assumptions is something that receives little or no mention in the teaching of
CLDs, and represents a challenge even for the seasoned practitioner. Especially significant are
the underlying assumptions of those working to develop the CLD, because they are not simple
errors in logic; they are unwitting errors in the input to the logical process. Even when the logic
of a CLD is impeccable, invalid assumptions underlying the diagram can render it ineffective or
even harmful, leading to disastrous interventions despite the best of intentions. Further, any
subsequent simulation model that preserves those assumptions will not reflect reality no matter
how precise or exhaustive the model might be.
Fortunately, some simple but potent tools make it possible to surface and challenge critical
assumptions underlying a CLD. This paper will describe the tools; present several easily
recognizable problems taken from the world around us, expressed in crude, first-draft CLD form;
and then apply the tools to illustrate their worthiness as well as the sometimes shocking effect
their use can have on the CLD development process.
Il. Meet the Tools
A. The Logical Approach
Few students today (in the US, at least) receive training in logic and rhetoric. While this may be
considered disastrous for any number of reasons, its relevance to the subject of this paper is that
those who have not been schooled in these areas are at a disadvantage when it comes to drawing
CLDs.
While it would be inappropriate to attempt to shoehorn an entire course on logic into this paper,
for our purposes we must at least review the category of logical fallacies known as "causal
fallacies" - since, by definition, every link in a CLD represents, at a minimum, an assumption of
causality.
Coincidental Correlation (post hoc ergo propter hoc)
A fallacy committed by assuming that, because event B followed event A (post hoc), therefore (ergo) B was
caused by A (propter hoc). (e.g., "Every morning the rooster across the street crows, and immediately after that
the sun rises. Therefore, the rooster must be causing the sun to rise by crowing.")
Joint Effect
Sometimes considered a special case of post hoc ergo propter hoc. A is held to cause B when in fact both are the
effect of a single underlying cause. (e.g., "The lifeguard noticed that, on days when bathers on his stretch of
beach bought significantly more ice cream than usual, he had to rescue significantly more swimmers in distress
than usual. He concluded that a combination of eating and swimming was unwise, just as his mother had always
warned him." The lifeguard neglected to observe that the amount of ice cream consumed and the number of
distressed swimmers both went up on days that were significantly hotter than usual - oppressive heat being the
underlying cause of both effects.)
Insignificant Cause
A fallacy committed when A, which does indeed contribute to B, is considered before more significant causes.
(e.g., "Smoking causes air pollution in New York City." This is true, of course, to a degree - but its effect is
insignificant compared to the effect of auto exhaust.)
Wrong Direction
A fallacy committed when the relationship between cause and effect is reversed. (e.g., "If we want our children
to receive a good education, we must make sure their grades are high." But high grades are not the cause of a
good education; rather, they are its result.)
Complex Cause
A fallacy committed when A, which does indeed contribute to B, is nevertheless only one of several causes that
are all necessary. (e.g., "The fireball that burst from my barbecue happened because I squirted lighter fluid onto
it." This statement is partially true, but without the presence of heat from the charcoal and oxygen in the
atmosphere, no fireball would have resulted.)
Figure 1: Causal Fallacies
B. In Search of Conflict
The Theory of Constraints, developed by Eli Goldratt, introduces a thinking tool called the
Conflict Resolution Diagram (CRD), as shown in Figure 2. The effectiveness of this tool rests on
the theory that problems within a system often exist due to powerful forces in conflict with one
another, usually due to incompatible underlying assumptions. The tool is designed to surface
those conflicts so they can be resolved.
Using the CRD requires serious thought, but the structure of the diagram itself is quite
straightforward. In order to achieve some stated objective, we must meet some requirements. In
order to meet those requirements, we must meet some prerequisites. Those prerequisites have
further prerequisites, and so on. If we continue this thought process, we will often encounter two
or more prerequisites that conflict with one another. Assumptions underlying that conflict, once
they have been exposed, can be invalidated or broken, leading to a breakthrough.
prerequisite A requirement A
objective
prerequisite B requirement B
Figure 2: Conflict Resolution Diagram
Ill. Apply the Tools
A. The Winning Athlete
1. Objective: High Level of Performance
Figure 3 illustrates the dynamics involved in a world-class figure skater's performance, which is
dependent on two factors: focus, meaning attention to a single task; and flow, meaning execution
of the task without explicit attention to the individual steps of the task. Self-consciousness and
anxiety about making mistakes are detrimental to focus and flow, and are therefore detrimental
to performance. Good performance reduces self-consciousness (R1) and anxiety (R2), while at
the same time raising expectations (B1). Poor performance, of course, has the opposite effect.
| . RI
Fr attention to task +>) focus N
self-consciousness a"
7 ~ performance
Pa level
4
expectations / Z
YS > 7, flow
+. \ . Zz
“ Echo attention to task |
anxiety about
ry at ts
making mistakes Someones
, R2
BI {
Figure 3: Achieving Focus and Flow
2. Chosen Means
This dynamic is familiar to coaches and sports psychologists, even if causal loop diagrams are
not. The most common interventions made on the athlete's behalf take the form of psychological
counseling and exercises (relaxation techniques, visualization, positive thinking, etc.) and
preparation (practice, conditioning, videotape reviews, etc.).
3. Discussion
While this could hardly be considered the finest CLD ever developed, it nonetheless captures the
essence of the situation and is an acceptable first draft. To begin improving it, let's introduce a
Conflict Resolution Diagram, as shown in Figure 4. Applied to this particular case, it illustrates
that if we trace back from each of the main requirements for high performance (focus and flow)
through a chain of prerequisites, we soon find that two of those prerequisites conflict with one
another.
high concern
over results
high desire for
success
high attention to task
(focus)
Jow attention to task
components (flow)
low concern low fear
over results of failure
Figure 4: Searching for Conflict is Not the Same as Looking for Trouble
Our awareness of the conflict helps us to see that this example is, at its core, a classic problem of
balancing risk with reward. If the reward is great enough, there will be a high concern over
results (equivalent to greater desire for success); and as the risk associated with failure is
reduced, there will be less concern over results (equivalent to lower fear of failure). To resolve
the conflict we must generate a concern over results that is high enough to challenge and
motivate, but low enough to encourage rather than intimidate - which many readers will
recognize as a parallel to a well-known strategy for effective goal-setting.
This leads us to the rather basic question of how the skater chooses goals and defines success and
failure - two words among several that we have been playing fast and loose with up to this point.
For any given competition, if winning that competition is considered the only acceptable
definition of "success," then the risk of failing to achieve that goal is extremely high. When
skating in exhibition, however, the skater has different goals and defines success and failure
quite differently than in competition; and "success" is far more likely, since that term is no
longer confined to a single narrowly defined outcome.
This thought process exposes an assumption implicit in our original approach (that
"performance," "winning," and "success" are identical) and suggests a slightly different
approach. The updated CLD in Figure 5 does not invalidate the interventions mentioned earlier,
but it does open the door to another family of interventions. For example, counseling can help
the skater recognize that a career as a competitive skater consists of many competitions, and
therefore consists of many opportunities to win beyond the current competition. Further, each
competition can be assigned one or more of a set of layered goals or objectives - improvement in
a particular jump or piece of footwork, increased artistry, greater speed, more endurance, and so
on - which will improve the skater's chances to win future events while also increasing the
number of successful outcomes available in any single competition as well as in a career of
competition.
attention to task
self-consciousness
JE
“4
expectations
»~
number of possible
‘outcomes that could be
labelled "success"
anxiety about
making mistakes
"
4
attention to task
components
flow
Figure 5: Objectives and Success
B. Holding Our Schools Accountable
1. Objective: High Standardized Test Scores
The standardized test scores of our nation's youth are the subject of much concern - and much
legislation. Standardized tests that had been mandatory only every other year between third and
eighth grade are now mandatory for every year in that range. Schools are being pressured to
increase standardized test scores, and the schools that fail to do so within the next few years are
threatened with being restructured (replacing all staff and faculty and revamping the curriculum)
or taken over by the state. By applying pressure to the staff and faculty of our schools in the form
of deadlines and reverse incentives, we hope to create accountability and ensure that test scores
rise.
2. Chosen Means
The dynamic is illustrated in Figure 6. Of course, most parties involved expect children who are
well schooled in the fundamentals to perform well on standardized tests (B2). As most parties are
also aware, however, the pressure to raise scores has generated some unintended side effects.
The pressure is severe enough that teachers begin "teaching to the test" - that is, preparing
students specifically for the tests - in hopes of raising the scores high enough and quickly enough
(B1). In so doing, they reduce the amount of time spent teaching fundamentals (R1).
teaching to the < Ea pressure to <
test improve scores
BI +
Ri
disappointing test
scores
time spent B2
specifically [
preparing for tests
teaching of
fundamentals, critical
thinking, problem
solving
Figure 6: Lighting a Fire
To solve this dilemma, some schools have set aside quotas of time specifically for test
preparation, with further reverse incentives designed to prevent teachers from exceeding those
quotas. The understanding is that this action will choke off loops B1 and R1, and force more
time into the desirable B2 loop - despite the loop's inherent delay and the looming deadlines.
3. Discussion
This dynamic is riddled with unexamined assumptions. Let's discover some of them using a
Conflict Resolution Diagram (Figure 7). First, we declare our objective: higher test scores. Given
our actions, one can infer that the requirements to achieve this goal are 1) more frequent
administration of the tests; and 2) pressure on schools to improve their scores.
increased
testing
better test
scores
pressure to
improve scores
Figure 7: Shooting for Higher Scores
At first it's difficult to see any cause-effect relationship between more frequent testing and
improved scores; but proponents of more frequent testing explain that, combined with the
pressure on schools to improve scores, it forms the accountability so necessary to bring about
improvement. The argument is not that the testing itself causes improvement, but that it provides
information for all parties to act upon in their improvement efforts - that is, it helps schools and
legislators to evaluate progress and identify which schools are failing and which are succeeding.
Consideration of the link between pressuring schools and better scores exposes one of the critical
assumptions underlying the dynamic. Since the pressuring legislation includes no fundamental
change in the system or means by which schools educate their charges, one of two things must be
assumed to be true of any school identified as a "failing" school: either the faculty and staff could
have been doing better all along but chose not to (and therefore need the special motivation that
threats will engender); or they are simply not capable of meeting the expectations placed upon it
(and therefore needs to be identified as such, for possible elimination). When the stakes are such
that the possibility exists of being not only unemployed, but implicitly labeled as lazy and/or
incompetent, is it any wonder that teachers might teach to the test that determines their fate, and
sometimes even ignore any restrictions on doing so?
This line of reasoning demands that we ask the question: Why don't teachers just do a better job
of teaching (the B2 loop) instead of teaching to the test? Given the way we have defined the
situation thus far, the answer again can only be one of two possibilities. Either they can't do it, or
they haven't been bothering to do it - and are now so far off course that they can't risk getting
back onto the B2 loop because they fear the inherent delay of that loop might cause them to miss
crucial deadlines.
The reader will recognize how grim this view of the situation is: some (unknown) number of
schools is either unable or unwilling to provide an acceptable education to our children; and
those schools even attempt to circumvent legislation intended to identify the incompetent and
motivate the unwilling, at further cost to our children's education. Fortunately, this assessment
rests on flawed assumptions; unfortunately, a more accurate assessment is even more grim.
If we take a moment to step back and consider the problem in a broader context, we can
acknowledge easily enough that the reason we want higher scores is that we want our children to
receive a better education. We want them to be well schooled in the fundamentals, able to think
for themselves, and prepared to deal with the myriad and unpredictable problems they will face
throughout their lives. Updating our previous CRD to include "better education" as our higher
goal (Figure 8) makes another assumption obvious: the assumption that the types of standardized
tests currently in use are appropriate, meaningful, and sufficient measures of educational
achievement. A bit of reflection reveals that this assumption is key to the entire dynamic. The
test results are the linchpin in the process of evaluating schools and determining the fate of
faculty and staff.
increased
testing
better test | 5 better
scores i > education
pressure to
improve scores
Figure 8: Aiming for a Higher Purpose
Consider the ramifications if the assumption were false, and the tests did not provide a useful
measure of the education being provided by the schools. The entire dynamic would crumble. All
improvement efforts that targeted the test scores would be seen as misguided. The labeling of
hundreds of schools as failures, and the termination and attendant stigmatization of thousands of
staff and faculty, would turn out to have been entirely unjustified, even libelous. Imagine the
destructive and demoralizing effects we would have visited upon an entire class of professionals
dedicated to the development of our youth.
A full determination of whether or not this assumption is flawed should be deferred to a separate
forum. Obviously, many believe that the tests accurately and reliably measure the quality of
education - at the same time, paradoxically, believing that the tests can be fooled by "teaching to
the test." The reader is invited to reflect on what compelling evidence exists for each side of the
argument, keeping in mind the stakes involved.
If we decide to accept the assumption that the tests are a useful measure of education, we
nevertheless still can see that the new connection in the CRD is a causal fallacy ("Wrong
Direction"). Good test scores do not lead to a good education; rather, a good education leads to
good test scores.
Let's start over with a new CLD (Figure 9) that now explicitly includes the objective of a quality
education, with the relationship between education and testing the right way around. Perhaps the
most striking effect of doing so is the way it focuses the CLD on process rather than results.
Instead of the pressure being on schools to produce better test scores, the pressure is on them to
do the things that ensure students receive a quality education. It follows that students should then
score better on any tests that measure education effectively.
The conclusions we may derive from this CLD are significantly different from our earlier
conclusions. Our previous diagram reflects the dynamic currently at work; our new one suggests
an entirely different - and more logically sound - approach. While testing remains useful
(assuming the tests are a good measure of education), there must be strong interventions focused
on measuring and improving the education process itself, not just on its results. This will require
of our leaders a great deal more work and a far deeper (and earlier) commitment than is required
to simply review test scores after the fact and point fingers of blame or hand out awards. To use
an automotive analogy: if we wish to travel the road of improving education for every child, then
looking at even meaningful test scores is like looking in the rear-view mirror at the road behind
us - a very small part of the process which, though advisable from time to time, gives no
indication of what lies ahead and makes no contribution to forward progress.
quality of
Pee us test scores —,
] Target
v
| BI gap
+
teaching of pressure to teach
fundamentals, critical fundamentals, critical
<t+ <—
thinking, problem
solving
thinking, problem
solving
Figure 9: The Quality of Education Is Not Strained
C. Reducing Industrial Accidents
1. Objective: Low Injury Rate
In any industrial setting, injuries caused by accidents are a serious source of concern. Their rates
are tracked, plotted, and pored over by supervisors, union leaders, regulatory agencies, and
countless others in offices everywhere, from the smallest factories to the largest industrial giants.
Reducing the injury rate, or keeping it low, is not just a humanitarian issue. It makes good
business and financial sense as well; insurance is cheaper, medical and leave expenses are lower,
and people feel better about working in a place they feel is safe.
2. Chosen Means
Perhaps the most well-known and often-practiced method of reducing injuries resulting from
industrial accidents is the use of safety incentive programs, the logic of which is illustrated in
Figure 10. Given a (presumably low) target for the number of injuries, incentive programs are
put in place to ensure that workers are properly informed and sufficiently motivated to behave
safely. Safe behavior reduces the number of injuries, shrinking the gap between the number of
injuries and the target.
Traditional safety incentive programs usually take one or both of the following forms:
e Performance incentives: rewards or prizes are given out, contingent on achieving or
maintaining some predetermined injury rate over some predetermined time period
e Participation incentives: rewards or prizes are given out, contingent on attending safety
awareness meetings, participating in safety-related activities, or successfully completing
safety training programs
=
number of injuries |) ——_—_+.
> Target
_ BI <i
ss a
- performance
-____ <<;
incentives
B2
participation
incentives
Figure 10: Avoiding the Dagger in the Floor
3. Discussion
It would appear that nothing could be more simple, elegant, or guaranteed of success than this
approach. Unfortunately, the key links between incentive programs and number of injuries suffer
from the causal fallacy of "Complex Cause."
Accident and injury rates are a function of risk, which can be avoided, but which also can be
reduced. Imagine a large dagger, its handle stuck in a factory floor and its blade pointing upward.
Workers in the factory can be warned of its presence and trained to avoid stepping on, stumbling
over, or falling onto it, and they can be rewarded for every month in which they have avoided
injury from it. Then again, it could simply be removed from the floor, removing the associated
risk along with it. In such a case, one might argue that the fallacy at work in Figure 10 is not
"Complex Cause," but "Insignificant Cause." Removing the dagger from the floor is a far more
powerful intervention than endless rounds of meetings, training, and awards programs.
Risk derives from multiple sources. First, workers might be unaware of safety hazards; second,
they might not know how to avoid or reduce them; third, they might not choose to avoid or
reduce them; fourth, the facilities and equipment might create risk; and fifth, some features of the
work environment might be counter-productive to the avoidance or reduction of risk. Traditional
safety incentive programs deal only with the first three sources (and only the avoidance
components of those), and not at all with the last two sources. Let's consider the impact of such a
strategy.
As seen in our previous example, imposing performance incentives based on results focuses
attention, as one might expect, on those results rather than on the process itself. Under the
circumstances, since the workers themselves are powerless to "remove the dagger," such
incentives encourage under-reporting of injuries (Figure 11) - actually contributing to the fifth
source of risk noted above. Some workers may even suspect that the company is trying to "buy
off" injuries to achieve lower (false) numbers. Further, these incentives send the demoralizing
message that the company believes the workers are foolish enough not to take heed of their own
safety unless there are trinkets to be had for doing so. Even worse, the company makes that
belief self-fulfilling and self-propagating by creating an addictive sense of entitlement - "If you
give me extra rewards just for not hurting myself, what extra reward will you give me for doing
my job and helping this company make money?"
number of injuries
>! , number of injuries | ——_4 rt i.
| reported | Target
- § p+ |
| » Bi v < |
} : | gap
. | performance |
B2 incentives | < +
participation
incentives *< ai
Figure 11: Safety Short-Circuited
In summary, such incentives demoralize the workforce, foster a sense of entitlement, distract
workers from even thinking about possible process improvements, and encourage the
falsification of vital injury data (all extremely negative consequences, extending beyond safety
issues into areas such as productivity, morale, retention, compensation, and trust). Yet they
remain popular, even despite the lack of any clear causal link between them and real injury rates.
Training in safe behaviors can be valuable, as can safety audits, and meetings to raise awareness
of safety issues; and if workers are concerned for their own safety they certainly will want to
participate in any activity that can help them behave safely. Additional incentives to participate
should be unnecessary. If they are, it should be clear that the workers do not find the activities
effective - and where there is no clear connection between the activity and an increased sense of
safety, the focus shifts once again to the rewards for participation and away from the substance
and content of the activity.
We have identified a causal fallacy in our original CLD - a fallacy that led to some sources of
risk being ignored - and we have also uncovered a familiar (by now) assumption that the workers
will not do what it takes to behave safely unless they are bribed (or threatened) into it. If we deny
that assumption and correct the causal fallacy, our next version of the CLD will look quite
different than it did before - perhaps something like Figure 12.
No one knows any better than the workers themselves where the "daggers" are in their
workspace. When they are engaged in identifying risk and in sharing meaningful knowledge of
risks with one another, and when they are given the authority to change facilities, equipment, and
their environment (physical and policy-related) to reduce risk and make remaining risks more
avoidable, real injury rates will decrease. It's not unreasonable to conclude that morale, trust, and
worker dedication might improve as well, not to mention productivity and pride of workmanship.
Of course, there wouldn't be quite so many trinkets gathering dust on shelves in workers’ homes;
but perhaps that is the price one must pay for a safer workplace.
improvement of
facilities and
equipment
removal of
factors
meaningful
risk-avoidance
engagement and
authority of workers in
risk identification,
reduction, and training
Figure 12: Building a Culture of Safety
risk reduction
+
counter-productive |
environmental risk avoidance
IV. Conclusion
There is much more that could be said about the diagrams discussed in this paper. No doubt, in
the course of our discussion there have been logical fallacies committed and critical assumptions
overlooked. Further effort will bring further revelations, and will require further revisions. The
"final" versions, if they may be called that, perhaps will bear only a passing resemblance to the
diagrams found here. It is certain, though, that they will be nothing at all like the earlier versions
described here, since the flaws in those earlier versions have been exposed.
The tools discussed in this paper do not eliminate the need for creativity or insightfulness; those
traits are still required among CLD developers. They are, however, useful weapons in a
developer's arsenal. They help developers by improving the process of developing CLDs; they
enable developers to think about problems in a more deliberate way; they contribute to making
underlying assumptions explicit; and they help to bring the powerful promise of CLDs to
fruition.
References
Blair, J. Anthony, and Johnson, Ralph Henry. 1994. Logical Self-Defense.
New York: McGraw-Hill.
Dettmer, H. William. 1997. Goldratt's Theory of Constraints.
Milwaukee: ASQC Quality Press.
Back to the Top
