A Comprehensive Model of Goal Dynamics in Organizations:
Setting, Evaluation and Revision’”
Yaman Barlas
Bogazici University, Dept. of Industrial Eng.
34342 Bebek, Istanbul — Turkey
Tel: +90 212 359 7073, Fax: +90 212 265 18 00
ybarlas@boun.edu.tr
Hakan Yasarcan
Accelerated Learning Laboratory
Australian School of Business (Incorporating the AGSM)
The University of New South Wales
Sydney, NSW 2052 — Australia
Tel: +61 2 9931 9193, Fax: +61 2 9931 9517
hakany@agsm.edu.au
Goal setting plays a central role in most simulation models of individual or social
behavior. In the simplest case, there is a constant goal and the modeling effort focuses on
the difficulties involved in reaching that given goal. In more realistic situations, the goal
itself is variable: it can erode as a result of various phenomena such as deeply rooted
traditions or frustration due to persistent failure, it can evolve further as a result of
confidence caused by consistent success, or it can be consciously evaluated and adjusted
periodically as a result of some formal process. In any case, 'goal dynamics' constitutes a
fundamental sub-problem in most situations dealing with dynamics of individual or social
behavior. As such, there has been considerable research effort on how to model the
dynamics of goal formation in system dynamics models. With respect to the three types of
goal dynamics mentioned above, in the literature there exist some model structures that
capture certain limited and linear 'goal erosion' dynamics. We extend the existing models
to obtain a most general theory of goal formation dynamics, by including performance
improvement capacity constraints, short term and long term time pressures in reaching the
set goal and more realistic, richer mechanisms of goal erosion. We show that the system
can exhibit very subtle non-linear problematic dynamics in such cases. The model is
generic in the sense that it offers a general theory of goal formation, including potential
goal erosion (caused by persistent poor performance) as well as positive goal evolution
dynamics (as a result of consistent success), and more complex dynamics resulting from
interactions of these two extremes. Our model also offers some adaptive goal setting
strategies to avoid the undesirable goal and performance erosion dynamics typically
experienced in complex, risky goal-seeking environments.
Keywords: goal seeking, goal dynamics, goal formation, goal revision, floating goals, goal
erosion, goal evolution, goal setting strategy
' Supported in part by Bogazici University Research Funds; grants no: 02R102 and 06HA305
? To appear in: Qudrat-Ullah, H., Spector, M., and Davidson, I. (Eds.). Complex Decision Making: Theory
and Practice, USA: Springer-Verlag, to be in print in 2007. (Reprinted for this conference with permission).
INTRODUCTION
Goal setting plays crucial role in decision making in organizations as well as in
individuals. Most improvement activities consist of the following cycle: set a goal,
measure and evaluate the current performance (against the set goal), take actions (e.g.
training) to improve performance, evaluate and revise the goal itself if necessary, again
measure and evaluate the performance against the current goal, and so on ... [Forrester
1975; Senge 1990; Lant 1992; Sterman 2000]. So goals constitute a base for the decisions
and the managerial actions. In an organization, the performance level is evaluated against a
goal and, further, the effectiveness of the goal itself can and must be periodically
evaluated.
Among various research methods to analyze the dynamics of goals in organizations (and
individuals), an important one is simulation modeling — more specifically system dynamics
modeling that is particularly suitable to model qualitative, intangible and ‘soft’ variables
involved in human and social systems [Forrester 1961; Forrester 1994; Morecroft and
Sterman 1994; Sterman 2000; Spector et al 2001]. System dynamics is designed
specifically to model, analyze and improve dynamic socio-economic and managerial
systems, using a feedback perspective. Dynamic strategic management problems are
modeled using mathematical equations and computer software and dynamic behavior of
model variables are obtained by using computer simulation [Forrester 1961; Ford 1999;
Sterman 2000; Barlas 2002]. The span of applications of the system dynamics field
includes: corporate planning and policy design, public management and policy, micro and
macro economic dynamics, educational problems, biological and medical modeling,
energy and the environment, and more [Forrester 1961; Roberts 1981; Senge 1990;
Morecroft & Sterman 1994; Ford 1999; Sterman 2000]. Since these problems are typically
managerial-policy oriented, structures that deal with goal dynamics play an important role
in most system dynamics models. It is therefore no surprise that modeling of goal
dynamics is an explicit research topic in system dynamics.
A fundamental notion and building block used in most policy models, is a ‘goal seeking’
structure that represents how a certain condition (or state) is managed so as to reach a
given ‘goal’ (see Figure | and 2 below). For instance, the state may be the inventory level
or delay in customer service, and the goals would be a set (optimal) inventory level or a
targeted service delay respectively. [For numerous examples see Sterman 2000 and
Forrester 1961]. Management would then take the necessary actions (Jmprovement in
Figure 1) so as to bring the states in question, closer to their set goal levels. The most
typical heuristic used to formulate Improvement is:
Improvement = (Goal — State)/SAT
where SAT, State adjustment time, is some time constant. The dynamics of this simplest
goal seeking structure is depicted in Figure 2: State approaches and reaches the Goal
gradually, in a negative exponential fashion. In more sophisticated and realistic goal-
seeking models, the goal is not fixed; it varies up and down depending on current
conditions, called a ‘floating goal’ structure [Sterman 2000; Senge 1990]. In this case, if
the performance of the system is persistently poor (in approaching the originally set goal),
then the system implicitly or deliberately lowers the goal (eroding goal). If, on the other
hand, the system exhibits a surprisingly good performance, then the goal may be pushed
further up (evolving goal). Another key component of a general goal setting structure is
expectation formation: Goals are set and then adjusted in part as a function of future
expectations [Forrester 1961; Sterman 1987]. This may involve the expectations of
management, expectations of participants, or typically both. Formulation of expectations is
a rich research and modeling topic with its roots in theories of cognitive research in policy
making [Spector and Davidsen 2000] and rationality; ranging from rational expectations to
satisficing and bounded rationality [Simon 1957; Morecroft 1983]. So, formulation of goal
setting, evaluation and seeking is a deep and important research topic in system dynamics
modeling.
In this paper we present a comprehensive goal dynamics model involving different types of
explicitly stated and implicit goals, expectation formation and potential goal erosion as
well as positive goal evolution dynamics. We further include a host of factors not
considered before, such as organizational capacity limits on performance improvement
rate, performance decay when there is no effort, time constraints, pressures and, motivation
and frustration effects. We show that the system can exhibit a variety of subtle problematic
dynamics in such a structure. The modeling setting assumed in the paper is an organization
in which a new 'performance' goal is set and a new program (e.g., a training activity) is
started to achieve the goal. In a service company, this may be a new training program set
‘to increase the customer satisfaction from 60% to 80% in one year' as measured by
customer surveys. Or in a public project, this may be a new educational program in a poor
neighborhood set ‘to increase the functional literacy rate from 80% to 90% in three years’
as measured by periodic tests. So there is a goal and there is also a time horizon set to
reach the goal. Note also that the environment described above implies that the Goal is
always approached from below and higher State levels mean always better for the system.
(As opposed to inventory management, where there is some ‘optimal’ target inventory
level so that the management increases the inventory if it is too low compared to the set
goal and lowers the inventory if it is too high).
The paper starts by elaborating on the simplest model of constant goal seeking dynamics,
by including a constraint on improvement capacity and a nominal decay rate out of the
state variable (Figure 3). We then gradually add a series of more realistic and complex
goal-related structures to the initial model. The purpose of each addition is to introduce and
discuss a new aspect of goal dynamics in increasingly realistic settings. In the first
enhancement, we introduce how the implicit goal in an organization may unconsciously
erode as result of strong past performance traditions. Next, we discuss under what
conditions recovery is possible after an initial phase of goal erosion. In the following
enhancement, we include the effects of time constraints on performance and goal
dynamics. Many improvement programs have explicitly stated time horizons and the
pressure (frustration) caused by an approaching time limit may be critical in the
performance of the program. We also show that in such an environment, the implicit short-
term self-evaluation horizons of the participants may be very critical in determining the
success of the improvement program. Finally we propose and test an adaptive goal
management policy designed to assure satisfactory goal achievement, by taking into
consideration the potential sources of failures discovered in the preceding simulation
experiments. We conclude with some observations on implementation issues and further
research suggestions.
THE SIMPLE GOAL SEEKING STRUCTURE AND ITS DYNAMIC BEHAVIOR
The simplest goal-seeking structure consists of a single fixed goal, a condition (called State
in Figure 1) that is managed and a management action (Jmprovement in Figure 1). Once the
goal is set, it is not challenged by the internal dynamics of the system or by any external
factor. As mentioned before, assuming a typical stock adjustment heuristic (i.e.
Improvement = (Goal — State)/SAT), this structure exhibits a pure negative exponential
goal-seeking behavior shown in Figure 2.
State
‘improvement
Goal
State adjustment time
Figure |. The simplest goal seeking structure (stock-flow diagram)
7 Goal 2 State E hprovemnens
: 1000
3
3 100
4) soo
3 so
a]
3 0
3 6
S00 50100 700,00 150.00 200.00
bace 1 Days
Figure 2. The simple goal seeking behavior generated by the model of Figure |
In our enhanced version of this simplest goal-seeking structure, we introduce two new
factors to make it more general and realistic:
i- If there is no improvement effort, State experiences a natural decay (Loss flow)
and,
ii- /mprovement rate is (naturally) limited by some Maximum capacity.
The outflow, Loss is assumed to be simply a fraction (/oss fraction) of the actual State.
This implies in essence that due to rapidly changing hi-tech organizational setting, the
performance level tends to decay over time, if no improvement effort is undertaken.
Proportional formulation means that as State goes up, so does the Loss rate. This is realistic
to some extent, but to prevent exaggeratedly high Loss rates, we place a limit on Loss as
well, called Maximum loss. (We choose Maximum loss to be less than Improvement
capacity, otherwise it would be impossible to ever fulfill the goa/).
Maximum capacity is constant in our models, as capacity management is beyond the scope
of this paper. The effective Improvement capacity, on the other hand is variable
(Accomplishment_motivation_effect x Maximum_capacity), but in this first model
Accomplishment motivation effect is equal to one, so it has no role yet. (It will have an
important role later in the enhanced versions of the model). In any case, /mprovement is
thus formulated as:
Improvement = MIN(Desired_improvement, Improvement_capacity)
The Desired improvement formulation is the standard ‘anchor-and-adjust’ formulation
[Sterman 1989; Sterman 2000; Barlas and Ozevin 2001; Yasarcan 2003]. It is given by:
Desired_improvement = Estimated_loss + State_adjustment
State_adjustment = (Stated_goal — State)/State_adjustment_time
So the Desired improvement decision uses Estimated loss as an anchor and then adjusts the
decision around it, depending on the discrepancy between the goal and the current state.
Since it is not possible to know the Loss immediately and exactly, it must first be estimated
by the decision maker or system participants. So, Estimated loss is the output stock (see
Figure 3) of an expectation formation structure, using simple exponential smoothing
formulation:
Estimation_formation = (Loss — Estimated_loss)/Estimation_formation_time
Estimation formation time represents the delay in learning the actual value of performance
Loss. All stock, flow and converter variables are shown in Figure 3.
State
Improvement Loss
Loss fraction
@
Improvement
capacity Maximum loss
Maximum
Accomplishment capacity
motivation effect
Estimation
formation time
Estimated loss
Estimation formation
ee Stated goal © eal goal
State
adjustment State adjustment time
Desired
improvement
Figure 3. Simple goal seeking model with improvement capacity limit and loss flow
Note that there are two different goal variables in Figure 3; Stated goal is set and declared
by the management, and Jdeal goal is defined as the best (highest) possible goal for the
system. In this first simple model, Stated goal is assumed to be equal to the /deal goal. (In
all simulation experiments, initial State is taken as 100 and Ideal goal as 1000. In the
simpler models and experiments, Stated performance goal is equal to Ideal goal, but
especially in more complex models, Stated goal will not be necessarily equal to the /deal
goal, and in the final model it will not even be constant).
4: Ideal goal 2: Stated goal 3: State
1 1000.
2
3
1
2 5004
3
1
2
3 0 r
0.00 50.00 100.00 150.00 200.00
IPage 1 Days
Figure 4. Goal seeking behavior generated by the model of Figure 3
(Stated_goal = 750)
1: Improvement capacity 2: Desired improvement 3: Improvement 4; Loss
IN
25:
, iw
2
3
a 0.
0.00 50.00 100.00 160.00 200.00
IPage 1 Days
Figure 5. Dynamics of flows related to the run in Figure 4
Dynamic behavior generated by simulating the model of Figure 3 can be seen in Figures 4
and 5. The behavior is a variant of standard ‘goal-seeking behavior’ seen in Figure 2. State
(line 3 in Figure 4) gradually seeks the Stated goal. Since the Improvement rate (line 3 in
Figure 5) is above the Loss rate (line 4 in Figure 5), the State level keeps increasing until it
reaches the Stated goal, at which point the improvement is lowered down to the Loss rate,
since the goal is reached. Also observe (in Figure 5) that when the Desired improvement
(line 2) is above the maximum improvement capacity (line 1), the actual improvement rate
stays constant at Maximum capacity. When Desired improvement is below the Maximum
capacity, then the actual improvement becomes equal to the Desired improvement.
This model is still too simple, being basically of introductory pedagogical value. The
model can explain simple goal seeking dynamics like heating of a room or a water tank
filling up to a desired level after flushing. In order to represent goal dynamics of human
systems and organizations, we incorporate a series of realistic enhancements in the
following sections.
TRADITIONAL PERFORMANCE, IMPLICIT GOAL AND EROSION
Goal erosion may occur if there is an endogenously created, undeclared, /mplicit goal that
system seeks, instead of the explicitly set goal (Stated goal). The model shown in Figure 6,
is more realistic and complex version of the simple goal seeking model, involving the
structures related to Implicit goal and eroding goal dynamics. Observe two important new
variables in Figure 6: Traditional performance and Implicit goal. Traditional performance
[Forrester 1975] represents an implicit, unconscious habit formation in the system. The
human element in the system gradually forms a belief (a self image) about his/her own
performance as time passes, and this learned performance (Traditional performance) may
start to have even more effect than the Stated goal [Forrester 1975; Senge 1990; Sterman
2000]. The accumulation of the individual beliefs creates a belief within the system that the
system can realistically perform around this past performance. To represent this
mechanism, we assume that the system creates its own internal goal called Jmplicit goal,
and seeks this new goal instead of the managerially Stated goal.
See. Loss fraction
Improvement Loss
@ ey i
©
Maximum loss
Improvement
capacity
Traditional
Maximum performance
capacity
Accomplishment
motivation effect
Estimated Traditional
loss performance
formation
Ideal goal
Desired
improvement Stated goal
Implicit goal
Weight of stated goal
Figure 6. A basic model of eroding goal dynamics
Observe that in the model of Figure 6, there are only two new variables compared to
Figure 3. The first one is Traditional performance which is essentially a historical
(moving) average of past performance (State). This is formulated by simple exponential
smoothing (with a tradition formation time of 30 days), just as it was done to formulate
Estimated loss, above. The second one is /mplicit goal, which is assumed to gradually tend
to the Traditional performance, starting with an initial value of Stated goal. (There is a
new parameter in Figure 6 called Weight of stated goal that should be ignored at this point,
as it has no role in this version of the model; this parameter will have a role in the
following section). So in a nutshell, Jmplicit goal is also an exponentially delayed function
of traditional performance:
Implicit_goal = SMTH3(Traditional_performance, 10, Stated_goal)
The third-order exponential smoothing function (SMTH3) used above means that the
output (Implicit goal) does not immediately react to a change in the input (Traditional
performance); there is a period of initial inertia. Implicit goal does erode to Traditional
performance, but the erosion should not start immediately and should not react to any
temporary change in traditional performance. Finally, the goal used in the Desired
improvement equation is now Implicit goal (instead of Stated goal).
We assume that in a new improvement program, in the beginning there is no concept of
past performance, so the human participants in the system completely accept the Stated
goal as their goal (i.e. Implicit_goal = Stated_goal). But as time passes, Traditional
performance starts to have bigger effect and the Implicit goal starts approaching the
Traditional performance instead of staying at the managerially Stated goal, a phenomenon
often called ‘goal erosion’. Note that together with this goal erosion, the State starts to
pursue the /mplicit goal, not the Stated goal, so the result of the improvement program is a
failure. The dynamic behaviors of goal erosion can be seen in Figure 7.
4: Stated goal 2: Implicit goal 3: State 4: Traditional performance
1 1000:
2
3
‘
1
2
2 5004
r
pon
1
2
3
4 °: t
0.00 2500 50.00 75.00 100.00
Page 1 Days
Figure 7. Strong erosion in goal, caused by Traditional performance bias
Goal erosion can be severe or mild, depending on some environmental factors. In the
scenario represented in Figure 7, we assume that the Traditional performance formation
time is relatively long (there is a strong past tradition), so the Jmplicit goal erodes toward
Traditional performance and may erode to the point of even crossing below the current
State level, since the highly delayed Traditional performance determines the Implicit
performance goal (see Figure 7). Erosion can be extreme if there is a very strong past
tradition. Conversely, if the tradition formation time is short, then goal erosion is milder
and also simpler: since with a short formation time, Traditional performance is almost
equal to current State, the Implicit goal would be effectively seeking the State (and the
State naturally seeking the Jmplicit goal).
GOAL EROSION AND RECOVERY
In the model shown in Figure 6, if the parameter Weight of stated goal is given a value
between 0 and 1, then the equation of /mplicit goal becomes:
Implicit_goal = SMTH3[Weight_of_stated_goal x Stated_goal +
(1-Weight_of_stated_goal)xTraditional_performance, 10]
The above equation states that Jmplicit goal is now basically a weighted average of the
Stated goal and Traditional performance. This weighted average is then passed through a
third-order smoothing to give it a realistic inertia, just as in the previous model. (But in the
previous version, Weight of stated goal was set to 0, so that Implicit goal was simply
exponentially eroding to Traditional performance, starting with an initial Stated goal). In
the current version of the model, participants are affected both by the external Stated goal
and by their Traditional performance, so that their Implicit goal is somewhere in between
the two extremes [Forrester 1961; Sterman 2000; Barlas and Yasarcan 2006]. Weight of
stated goal determines how much the system believes in Stated goal.
1: Stated goal 2: Implicit goal 3: State 4; Traditional performance
aa
\\
500:
0.00 62.50 125.00 187.50 250.00
IPage 1 Days
Figure 8. Behavior of goal erosion and recovery model (Weight_of_stated_goal = 0.5)
The resulting dynamics with Weight_of_stated_goal = 0.5 are shown in Figure 8. After a
significant initial erosion, Jmplicit goal and hence State gradually recover towards Stated
goal. Observe that since Traditional performance is an exponential average of past State
values, the former moves - fast or slow- towards the State after some delay. On the other
hand, the other component of the weighted average, Stated goal is fixed. The net result is
that all variables, including Traditional performance eventually recover and gradually
approach Stated goal over time (Figure 8). Thus, although this model is more realistic than
the previous version in some sense, it suffers from the following fundamental weakness:
the formulation implies that, even though a system may suffer from initial goal and
performance erosion, in time it will always recover (fast or slow) and attain the Stated
goal. The weakness is that there is no notion of ‘time horizon’ and potential frustration (or
motivation) to be experienced by the system participants having evaluated their
performances against the time constraints. In reality, Implicit performance goal may
continue to erode if there is a belief in the system that the set goal (Stated goal) is too high
or impossible to reach in some given time horizon. These concepts are introduced in the
following enhancements.
TIME HORIZON EFFECTS: FRUSTRATION, MOTIVATION AND POSSIBLE
RECOVERY
The two essential enhancements in this model are the concepts of project Time horizon and
accomplishment motivation (or frustration) that results from the participants’ assessment of
their performance against this time horizon. The project time horizon is taken as 200 days
in the following simulation runs. Time horizon is a stock variable representing how many
days left to the end of the project, starting at the initial horizon value, and depleting day by
day (see Figure 9). One subtle feature of this Time horizon stock is that it does not quite
deplete to 0; it stops when it reaches what we call ‘Short term horizon’ taken as 12 days.
The idea is that if the goal is reachable in just another 12 days, these twelve days are
always allowed. (The Short term horizon will have an active role in the next enhancements
and will be discussed in the following section). So the depletion rate of Time horizon is
given by:
Time_horizon_depletion_rate =
IF Time_horizon > Short_term_horizon THEN | ELSE 0
Accomplishment motivation effect is a factor that represents the participants’ belief that the
stated performance goal is achievable within the Time horizon. To formulate this, we
represent Accomplishment motivation effect as a decreasing function of Remaining work
and time ratio as shown in Figure 10. Remaining work and time ratio is an estimate of how
many days would be needed to close the gap between the current State and Stated goal,
relative to the remaining Time horizon:
Remaining_work_and_time_ratio =
[(Stated_goal — Perceived_performance)/5]/Time_horizon
In the above formulation, the discrepancy between Stated goal and Perceived performance
is first divided by the maximum rate at which State can be improved (which is
Maximum_capacity — Maximum_loss = 15 — 10 = 5), yielding how many days would be
needed at least to close the gap. (Perceived performance is just an exponentially smoothed
average of State). The result is then divided by remaining Time horizon to provide a
normalized ratio. (This ratio is also smoothed in the model, so that motivation does not
change too fast, without any inertia). Figure 10 states that motivation is full (equal to one)
when the above ratio is less than or equal to 0.7, it starts dropping afterwards and when the
ratio becomes about 2, it denotes complete frustration (equal to zero).
Accomplishment motivation effect plays two different roles in the model. First, this
motivation increases the effective improvement capacity of the participants:
Improvement_capacity = Accomplishment_motivation_effect x
Maximum_capacity
where Maximum _capacity = 15
Thus, the more motivated the participants are, the closer becomes the actual Jmprovement
capacity to Maximum capacity. (In the earlier versions of the model, Accomplishment
motivation effect was set to 1, so it had no effect). This first role of Accomplishment
motivation effect is represented by the positive feedback loop shown in Figure 11). The
second role of Accomplishment motivation effect is to influence Weight of stated goal as
follows:
Weight_of_stated_goal = Accomplishment_motivation xReference_weight
where Reference_weight = 1.0
See Loss fraction
Improvement Loss
@ — a)
Improvement Maximum loss
capacity
Ma: Traditional
ay performance
capacity
Accomplishment Traditional
motivation effect Estimated performance
loss: formation
Ideal goal
Desired
improvement
, Implicit
Perceived eal eee Short term horizon
performance
Time horizon
Weight of
stated goal
Remaining work Time horizon
and time ratio depletion rate
Figure 9. A model of goal erosion and possible recovery with a time horizon
Thus, Weight of stated goal is now a variable, depending on the motivations of the
participants. When participants have full motivation (1.0), then Weight of stated goal
becomes 1.0 as well, so that in forming their /mplicit goal, participants give 100% weight
to Stated goal and no weight at all to Traditional performance. At the other extreme, with
zero motivation, all weight is given to Traditional performance and zero weight to Stated
goal (meaning complete goal erosion). In between these two extremes, some non-zero
weights are given both to Stated goal and to Traditional performance, depending on the
level of motivation (see Figure 10).
1: Accomplishment motivation effect
1 1.00
0.00 0.50 4.00 1.50 2.00
IPage 1 Remaining work and time ratio
Figure 10. Accomplishment motivation effect is a function of
Remaining_work_and_time_ratio
aneroveriont rs
Implicit goal
i
7 Es
g
Desired Porcehved
improvement performance
Improvement
capacity (+)
a Stated goal
Aecomptsiment
ee ee
Time reer
Figure 11. The State seeks the Implicit goal, via Improvement (the basic inner goal-
seeking loop). Simultaneously, the Perceived performance relative to Time horizon
determines the Accomplishment motivation effect which in turn affects the Improvement
(the outer reinforcing loop).
Two typical dynamics generated by this model are depicted in Figures 12 and 13. In both
dynamics, there is an initial phase of strong erosion in J/mplicit goal, because the Stated
goal level is too high compared to the initial State (hence Traditional performance). After
this initial erosion, the next phase is one of recovery: Implicit goal starts moving up
gradually, and thus pulling up the State and Traditional performance. Finally, there is a
third phase in the dynamics that is interesting: In Figure 12, all three variables, after having
improved significantly, reverse their patterns and a final phase of erosion begins,
continuing all the way to the end. The mechanism behind this final erosion is related to the
time horizon of the project and the negative effects of de-motivation resulting from the
impossibility of reaching the goal, given the time constraint. This is observed in Figure 12,
where Stated goal is set to 1000 and Time horizon at 200. In Figure 13, on the other hand,
Stated goal is set at a lower value of 750 and the second phase of erosion never takes
place. The reason why all variables (Jmplicit goal, State and Traditional goal) continue to
improve is that Stated goal is now set at a more ‘realistic’ value, relative to the given Time
horizon. So the gap between Stated goal and Perceived performance relative to the
remaining time horizon never becomes so high as to cause a hopeless situation for the
participants. At the end, State can sustain a value of 750, which is below the /deal goal of
1000, but much better than the ‘giving up’ dynamics of Figure 12.
41: Stated goal 2: Implicit goal 3: State 4; Traditional perf ormance
1000-45 4 4 4
po
ET SN
500:
0.00 75.00 160.00 228.00 300.00
IPage 1 Days
Figure 12. Erosion dynamics, when Stated goal is too high for the given Time horizon
(Stated_goal = 1000, Time_horizony = 200)
1: Stated goal 2: Implicit goal 3: State 4: Traditional perf ormance
1000
|
0.00 75.00 160.00 228.00 300.00
IPage 1 Days
Figure 13. Erosion-then-recovery, when Stated goal is low enough for the given Time
horizon (Stated_goal = 750, Time_horizono = 200)
ROLE OF SHORT-TERM HORIZON IN POTENTIAL RECOVERY
A second component of motivation may have to do with the participants’ own intrinsic
Short term horizon, (assumed to be 12 days). The assumption is that participants judge
their own performance over a 12-day horizon and if they are satisfied (high sort term
motivation), then they give more weight to the Stated goal, otherwise they lower this
weight (meaning that the weight of their Traditional performance increases).
State
{) Loss fraction
Improvement
@
{_) Maximum loss
& Improvement
capacity
Traditional performance
Traditional
performance
formation
Accomplishment e~p
‘motivation effect
Desired
improvement
Short term
‘motivation effect,
Short term|
horizon
Short term
work and time ratio
Time horizon
4 goal
Remaining work Time horizon
and time ratio depletion rate
Figure 14. Time horizon, Short term horizon and their corresponding motivation effects
The new equation for Weight of stated goal becomes:
Weight_of_stated_goal =
Accomplishment_motivation xShort_term_motivation xReference_weight
where Reference_weight = 1.0
The formulation of Short term motivation effect is very similar to that of long term
Accomplishment motivation effect described in the previous section: The motivation is a
decreasing (from | down to 0) function of Short term work and time ratio, defined just like
Remaining work and time ratio defined above, except that the ratio is divided by Short
term horizon (12 days constant) instead of Time horizon of the project. One implication is
that if participants perceive that only 12 days of full effort can take them to the goal, then
they never give up, even if one day is left to the project Time horizon. Thus, the weight of
Stated goal (hence potential erosion of /mplicit goal) depends more subtly on the dynamics
of both the short and long term accomplishment motivations of the participants. These long
term and short term motivation interactions and main loops are shown in figure 15.
State + Short term
a Het horizon
Perceived
¥
Improvement performance
Traditional + fe
performance Short term
Implicit goal ” motivation effect
. -
+ Accomplishment
motivation effect
7
- i
(1 - Weight of SG)-(Traditional performance) Time horizon
a Stated goal
Weight of
(Weight of SG)(Stated goal) <j—————_ Stated goal +
3
Figure 15. A weighted average of Traditional performance and Stated goal determines
Implicit goal after some delay. The weight depends on the short term and long term
accomplishment motivations, which in turn depend on perceived remaining performance
gaps (relative to the time horizons)
As will be seen below, where the Stated goal is set turns out to be critically important in
this new model. The dynamics for the different values of Stated goal are plotted in Figures
16, 17 and 18.
4: Stated goal 2: Implicit goal 3: State 4: Traditional perf ormance
1 600
2
3:
4
1
2
Fi 300:
4
4 =
1
2
3:
4 o r =
0.00 75.00 180.00 225.00 300.00
IPage 1 Days
Figure 16. After an initial erosion in /mplicit goal, long term stagnating State eventually
results in giving-up behavior (Stated_goal = 600;
Time_horizong = 200)
If the Stated goal is too high (for instance 600) relative to the initial State (100) and Time
horizon (200), there develops a disbelief in the system that the stated goal is ever
reachable. This disbelief results in de-motivation, which further causes the Weight of stated
goal to reduce to near zero and the /mplicit goal erodes (Figure 16). After this initial
eroding goal behavior, motivations never become high enough to ignite a performance
improvement, but they can at least sustain the performance at some level for some duration
of time. Finally, the system participants recognize that the remaining time to accomplish
the Stated goal is impossibly too short, which results in ‘giving-up’ behavior and
improvement activity eventually dies down (Figure 16).
4: Stated goal 2: Implicit goal 3: State 4: Traditional performance
600:
\
300-4
Lae
0.00 75.00 180.00 228.00 300.00
IPage 1 Days
Figure 17. Initial erosion, then recovery and finally giving-up behavior due to time limit
(Stated_goal = 450; Time_horizono = 200)
The next simulation experiment starts with a lower Stated goal. In Figure 17, firstly we
observe an eroding goal behavior in the short term, due to the initial gap between Stated
goal and Traditional performance. But because Stated goal is not too high (450), this time
the Weight of stated goal does not become too small, which allows a goal recovery phase
between days 50 and 175. But, still the Stated goal is not low enough to prevent a giving-
up behavior in longer term due to the time limit. So, in Figure 17, three different stages of
goal dynamics can be observed: initial goal erosion, goal (and state) recovery, and finally
giving-up behavior.
1: Stated goal 2: Implicit goal 3: State 4; Traditional performance
600:
a
300:
\
wet
0.00 75.00 160.00 228.00 300.00
IPage 1 Days
Figure 18. If Stated goal is low enough, sustainable recovery is possible
(Stated_goal = 400; Time_horizono = 200)
Finally in Figure 18, Stated goal is low enough (400) to create a success: the recovery is
sustained and the goal is reached in the given Time horizon. The three dynamics (Figures
16, 17 and 18) show that if Stated goal is low enough relative to initial State and Time
horizon so as to create enough motivation, the State will improve towards the goal and
achieve it.
But, how can a manager know the ‘correct level’ of the Stated goal? Furthermore, what if
this level of performance is too low (conservative) compared to the true potential of the
system participants? Our comprehensive model and simulation experiments so far illustrate
these issues. As a result, we conclude that solutions to the above problems necessitate use
of a “dynamic and adaptive goal setting” strategies by the management, as will be
addressed in the next section.
ADAPTIVE GOAL SETTING POLICY FOR CONSISTENT IMPROVEMENT
J. W. Forrester (Forrester 1975) states: “...The goal setting is then followed by the design
of actions which intuition suggests will reach the goal. Several traps lie within this
procedure. First, there is no way of determining that the goal is possible. Second, there is
no way of determining that the goal has not been set too low and that the system might be
able to perform far better. Third, there is no way to be sure that the planned actions will
move the system toward the goal.”
State
Min acceptable
improvement rate
Time horizon
&_), Manager's operating horizon
() Goal achievable by trend
Ps
Reference level
formation time
Min acceptable (J
goal
Reference
level
(D Meal goal
{_) SG formation time
CO stated goal
Figure 19. Proposed adaptive dynamic Stated goal setting policy structure
In order to address these uncertainties, Stated goal should be set and managed dynamically
and adaptively. The management must continuously monitor the Stafe and must evaluate
its level and its trend. The Stated goal should be then set realistically within the bounds of
a “reachable region”, which is a function of the State /evel and its trend (net improvement
rate). If State is improving, Stated goal must also be gradually moved up to guide and
motivate the improvement activities. On the other hand if the State is stagnating, this
means that Stated goal is unrealistically high, so it must be lowered till there is a sign of
sufficient improvement in the State level. The structure in Figure 19 is designed to
implement and test such an adaptive management strategy.
In the related formulations, we assume that top management does not know Jmplicit goal,
Weight of stated goal, the motivation effects, Short time horizon, Capacity and Loss flow.
Management can only perceive the system performance (State) over time, so the Stated
goal decisions must be based on this information. If State is not improving enough, Stated
goal should be lowered, and if State is improving then Stated goal must also be gradually
moved up. Beyond this, exactly how much Stated goal should be moved up or down,
depends on several factors: Stated goal can not be bigger than /deal goal and it should not
be lower than some minimum acceptable goal determined by the top management. If the
level determined by the trend in State is in acceptable region, then Stated goal must be
equal to this level:
Stated_goal =
MIN(Ideal_goal, MAX(Goal_achievable_by_trend, Min_acceptable_goal))
Ideal goal is a given constant (1000) as discussed earlier. The second variable, Goal
achievable by trend is a managerial estimate of the current improvement trend and what
level can be attained at this rate, in some time horizon:
Goal_achievable_by_trend = State +
((State — Reference_level)/Reference_level_formation_time) x
Manager's_operating_horizon
Reference_level = SMTH3 (State, Reference_level_formation_time)
Manager's_operating_horizon = MIN(90, Time_horizon)
In the above formulation, Reference level is an average of past State, used in estimating the
trend, by dividing the improvement by Reference level formation time. Manager's
operating horizon is used in extrapolating the trend into the future. This horizon is equal to
the time horizon of the project as the project advances. But in early phases, the managerial
horizon is set to a smaller value (90), because it is assumed that extrapolating the current
trend farther into the future would be too uncertain. Finally, the third component of Stated
goal equation is some minimum acceptable goal determined by the top management. This
Min acceptable goal is determined by adding some minimum acceptable improvement rate
(in Manager's operating horizon) on top of the current State. It is assumed that a given
management has some minimum acceptable improvement rate, below which is simply
unacceptable:
Min_acceptable_goal =
State + Min_acceptable_improvement_rate xManager's_operating_horizon
Min_acceptable_improvement_rate = 0.5
In the above formulation, managerial constants like Manager's operating horizon, Min
acceptable improvement rate, and Reference level formation time are set at some
reasonable values that serve our research purpose. In an actual study, these constants must
be well estimated by data analysis and interviews and also tested by sensitivity analysis.
The adaptive goal setting policy structure is integrated in the full model (of Figure 14) and
simulation experiments are run under different scenarios. In all scenarios, it is assumed that
management starts with a rather high initial Stated goal (900), to demonstrate the fact that
the starting stated goal is not important anymore, because Stated goal is a variable in the
adaptive goal setting policy. In figure 20, the provided Time horizon is quite short (200), so
the expected behavior, based on previous experiments is one of strong erosion and give-up
behavior towards the end (for instance, Figures 16 and 12). But the dynamics in Figure 20
display an obvious improvement: Initially, Stated goal is deliberately lowered (with some
oscillations) by top management so as to ignite participant motivation, and then later it is
moved up gradually and adaptively, pulling together with it the /mplicit goal and State. At
the end, although reaching the ideal goal of 1000 was impossible in the given Time
horizon, State has improved to a reasonable level (500) within the time limits, without
displaying any giving-up behavior.
In the second experiment, a longer Time horizon (350) is provided. The main dynamic
characteristics are the same as those observed in the previous run: Stated goal is
deliberately lowered initially by top management, and then later, it is moved up gradually
and adaptively, pulling together the /mplicit goal and State (Figure 21). At the end, since
the Time horizon is longer, State is improved to a higher level (750), compared to the
previous run (although still lower than the ideal goal of 1000). From these two runs, the
important contribution of the adaptive goal setting policy is obvious: The performance
consistently improves and eventually settles down at a level without any giving-up
behavior at the end, which apparently is a strong improvement within the given Time
horizon.
1: Stated goal 2: Implicit goal 3: State 4; Traditional performance
1000
500:
aa
\\
0.00 62.50 125.00 187.50 250.00
IPage 1 Days
Figure 20. A satisfactory result with the dynamic goal management policy, even when the
initial Stated goal is high (900) and
Time horizono is short (200)
1: Stated goal 2: Implicit goal 3: State 4; Traditional performance
1 1000
2
3
i
os Sa |
1 Lye
t yo | es -
: on eee
; [
8
2
| eo
1
2
3
4 .
3.00 100.00 700.00 300.00 7400.00
pave 1 Days
Figure 21. A better result with dynamic goal management policy, when more time is
provided (initial Stated_goal = 900 & Time_horizong = 350)
Finally, when Time horizon is sufficiently long relative to the /deal goal of the project, the
last simulation experiment demonstrates that /deal goal (1000) can be reached. In Figure
22, Time horizon is taken as 500 and we observe that with sufficient time, the /deal goal is
eventually attained within the program Time horizon, implying maximum organizational
success.
1: Stated goal 2: Implicit goal 3: State 4; Traditional perf ormance
1000. OE <I
i LE
500: =
0.00 137.50 275.00 412.50 850.00,
IPage 1 Days
Figure 22. A close optimal result: /deal goal is attained via dynamic goal management,
when Time horizon is long enough (initial Stated_goal = 900 & Time_horizony = 500)
CONCLUSIONS
In the simplest computer/simulation models of goal-seeking in organizations, there is a
constant goal and the model describes the dynamic difficulties involved in reaching that
given goal. In more sophisticated models, the goal itself is variable: it can erode as a result
of various phenomena such as frustration due to persistent failure or it can evolve further
as a result of confidence caused by success. There exist some models of limited and linear
goal erosion dynamics in the literature. We extend the existing models to obtain a
comprehensive model of goal dynamics, involving different types of explicitly stated and
implicit goals, expectation formation and potential goal erosion as well as positive goal
evolution dynamics. The model constitutes a general theory of goal dynamics in
organizations; involving a host of factors not considered before, such as organizational
capacity limits on performance improvement rate, performance decay when there is no
improvement effort, time constraints, pressures, and motivation and frustration effects. We
show that the system can exhibit a variety of subtle problematic dynamics in such a
structure.
We build a series of more and more realistic and complex goal-related structures. The
purpose of each enhancement is to introduce and discuss a new aspect of goal dynamics in
increasingly realistic settings. In the first enhancement, we introduce how the implicit goal
in an organization may unconsciously erode as result of strong past performance traditions.
Next, we discuss under what conditions recovery is possible after an initial phase of goal
erosion. In the following enhancement, we include the effects of time constraints on
performance and goal dynamics. We also show that in such an environment, the implicit
short-term self-evaluation horizons of the participants may be very critical in determining
the success of the improvement program. Finally we propose and test an adaptive goal
management policy that is designed to assure satisfactory goal achievement, by taking into
consideration the potential sources of failures discovered in the preceding simulation
experiments.
Our theoretical model and management strategies can be implemented to specific
improvement program settings, by proper adaptation of the model structures and
calibration of parameters. Several managerial parameters in our models are set at some
reasonable values that serve our research purpose. In further applied research and actual
studies, these constants must be well estimated by data analysis and interviews and also
tested by sensitivity analysis. Our models can also be turned into interactive simulation
games, microworlds and larger learning laboratories so as to provide a platform for
organizational learning programs. More generally, our models may provide useful starting
points for different research projects on goal setting, performance measurement, evaluation
and improvement.
REFERENCES
Barlas, Y., 2002, System Dynamics: Systemic Feedback Modeling for Policy Analysis,
in Knowledge for Sustainable Development: An Insight into the Encyclopedia of Life
Support Systems, UNESCO Publishing — Eolss Publishers (Paris, France; Oxford, UK), pp.
1131-1175.
Barlas, Y. and M. G. Ozevin, 2004, “Analysis of Stock Management Gaming
Experiments and Alternative Ordering Formulations”, Systems Research & Behavioral
Science, Vol. 21, Issue 4, pp. 439-470.
Barlas, Y. and H. Yasarcan. 2006, Goal Setting, Evaluation, Learning and Revision: A
Dynamic Modeling Approach, Evaluation and Program Planning, Vol. 29, Issue 1, pp 79-
87.
Ericsson, K. A., 2001, Attaining Excellence Through Deliberate Practice: Insights
from the Study of Expert Performance, in M. Ferrari (ed.), The Pursuit of Excellence in
Education, Erlbaum (Mahwah, NJ), pp. 21- 55.
Ford, A., 1999, Modeling the Environment. Island Press (Washington, DC).
Forrester, J. W., 1961, Industrial Dynamics, MIT Press (Massachusetts).
Forrester, J. W., 1975,. Planning and Goal Creation: Excerpts, in Collected Papers of
Jay Forrester, Wright-Allen Press (Massachusetts), pp. 167-174.
Forrester, J. W., 1994, Policies, Decisions, and Information Sources for Modeling, in J.
D. W. Morecroft & J. D. Sterman (eds.), Modeling for Learning Organizations,
Productivity Press (Portland, Ore.), pp 51-84.
Lant, T. K., 1992, Aspiration Level Adaptation: An Empirical Exploration,
Management Science, Vol. 38, Issue 5, pp. 623-644.
Lant, T. K., & Mezias, S. J., 1992, An Organizational Learning Model of Convergence
and Reorientation, Organization Science, Vol. 3, pp.47-71, (reprinted in Cohen, M. D. &
Sproull, L.S. (eds.), 1995, Organizational Learning, Sage Publications).
Morecroft, J. D. W., 1983, System Dynamics: Portraying bounded rationality. OMEGA
The International Journal of Management Science, Vol. 11, Issue 2, pp. 131-142.
Morecroft, J. D. W., & Sterman, J. D. (eds.), 1994, Modeling for Learning
Organizations, Productivity Press (Portland, Ore.).
Newcomer, K., & Neill, S. (Eds.), 2001, Assessing Program Performance:
Measurement and Evaluation, special issue of Evaluation and Program Planning, Vol. 24,
Issue 1.
Roberts, E. W., 1981, Managerial Applications of System Dynamics, MIT Press
(Cambridge, Massachusetts).
Schalock, R. L., & Bonham, G. S., 2003, Measuring Outcomes and Managing for
Results, Evaluation and Program Planning, Vol. 26, Issue 3, pp. 229-235.
Senge, P. M., 1990, The Fifth Discipline: The Art and Practice of the Learning
Organization, Currency and Doubleday (New York).
Simon, H., 1957, Administrative Behavior: A Study of Decision Making Processes in
Administrative Organizations, Macmillan (New York).
Spector, J. M., & Davidsen, P. I., 2000, Cognitive Complexity in Decision Making and
Policy Formulation: A System Dynamics Perspective, in R. Sanchez & A. Henee, Selected
Papers from the 4th International Conference on Competency Based Management.
Spector, J. M., Christensen, D. L., Sioutine, A. V., & McCormack, D., 2001, Models
and Simulations for Learning in Complex Domains: Using Causal Loop Diagrams for
Assessment and Evaluation, Computers in Human Behavior, Vol. 17, pp. 517-545.
Sterman, J. D., 1987, Expectation Formation in Behavioral Simulation Models,
Behavioral Science, Vol. 32, pp. 190-211.
Sterman, J. D., 1989, Modeling Managerial Behavior: Misperceptions of Feedback in a
Dynamic Decision Making Experiment, Management Science, Vol. 35, No. 3, pp. 321-
339.
Sterman, J. D., 2000, Business Dynamics: Systems Thinking and Modeling for a
Complex World, McGraw-Hill (New York).
Yasarcan, H., 2003, Feedback, Delays and Non-linearities in Decision Structures,
Ph.D. Thesis, Bogazici University.
Yasarcan, H., & Barlas Y., 2005, A Generalized Stock Control Formulation for Stock
Management Problems Involving Composite Delays and Secondary Stocks. System
Dynamics Review, Vol. 21, Issue 1, pp. 33-68.
