Give me the right goals, I will be a good dynamic decision-maker
Jenshou Yang
Department of Business Administration, National Yunlin Institute of Technology, 123,
University Road Section 3, Touliu, Yunlin, Taiwan, R.O.C.; E-mail: yangjs@ba.yuntech.edu.tw
Introduction
Previous studies on dynamic decision making found people fail to control dynamic complex
tasks. A number of bounded rationality in dynamic decision making were found including (a)
narrow time span of thinking where ignoring time delay (Sterman, 1989), (b) unable to dealing
with interdependence among subsystems (Forrester, 1995), (c) linear decision rule which is not
appropriate for nonlinear characteristic of dynamic complexity (Doerner, 1980). Given the
bounded rationality of dynamic complexity, this study focuses on how to improve dynamic
decision performance through goal setting.
A Systems Archetype View of Poor Dynamic Decision Performance
Subjects perform worse because they are bounded rational on dynamic complexity and thus
trapped by some unrecognized systems archetype. For example, it is the Beer Game that belongs
to Balancing with Delay structure, and the minimize stock/backlog goal thereby forcing
participants to overshoot (see Figure 1). Goals dictate people's behavior and are the necessary
factor to produce overshooting behavior. The Balancing with Delay structure is the other
necessary condition which results in overshooting behavior.
-
order oo,
\ side effects
loop R1 pressure to + goal to
deli oe backlog
livery log
arrival delay + +
licy loop B1
poley oP backlog
inventory -
Figure 1. Beer Game is a Balancing with Delay systems archetype
The task adopted by Sterman (1989) is another case. As shown in Figure 2, the structure
combine the "minimizing gap goal" to be a Fix That Fail systems archetype. Large amount of
order fluctuation can not be produced by only subjects’ unrecognition of unexpected reinforcing
loop R1 unless they accepted the goal of minimizing the gap.
ZY production ~e
capital licy loop B1 Sap wet goal to
stock Peeyee minimize gap
: +
+ orderey, [prcssureitg
Forder +—“‘otal backlog
side effects loop R1
+ backlog from *
. Capital sector
Figure 2. The task used by Sterman (1989) is a Fix That Fail systems archetype
709
Yang (1996) found when subjects accepted 30% order growth rate goal they suffered
Growth and Underinvestment behavior as Growth and Underinvestment systems archetype
predicted. As shown in Figure 3, the combined structure including "maximize order growth
goal" became a Growth and Underinvestment systems archetype. Given subjects’ bounded
rationality of dynamic complexity, they hired too many salesmen and hesitated to invest resulting
in underinvestment and delivery delay. Poor performance occurred. Underinvestment can not be
produced by only subjects’ unrecognition of unexpected balancing loop B1 unless they accepted
the maximizing order growth goal.
pressure to hire <«t_ goal to maximize order growth
+
tog salesman +
+ ai A. A
hiring RI Bl delivery capacity
we order delay B2 l
+ revenue eZ \ ed - i +
Figure 3. The task used by Yang (1996) is a Growth and Underinvestment systems archetype
Improving Performance via Setting Right Goals
Given the proposition that bounded rationality of dynamic complexity, counterintuitive
structure, and the goals combine to determine some dysfunctional systems archetype behavior
and poor performance. The hypothesis of this study was dynamic decision performance could be
improved via right goal settings.
Method
Task
The dynamic decision task, a management flight simulator, as shown in Figure 4, was a
simulated ecosystem. The decisions in the task were the number of prey hunting and predator
killing. Subjects managed the system via the two decisions to reach the assigned goals.
Figure 4 Causal structure of the task
Design
Three kinds of goal setting were manipulated as following under the consideration that the
number of prey and predator and the prey/predator ratio was 1400, 492, and 2.85 respectively at
steady state. Ten percent of variation was allowed. The first was prey/predator ratio goal, a
710
wholesystem ratio goal, in which subjects were instructed to maintain the ratio between 2.57 and
3.14. The second was prey/predator number goal, a wholesystem number goal, in which
subjects were instructed to maintain the number of prey within 1260-1540 and the number of
predator within 443-541 at the same time. The third was prey number goal, a subsystem goal, in
which subjects were instructed to maintain the number of prey within 1260-1540. The more
times subjects reached the goals, the better their performance.
Dependent Variables
Fix that fail decisions Fix That Fail dysfunctional behavior occured when subjects were
bound rationally ignoring the side effects of hunting or killing too many prey or predator to reach
the goals. The hypothesis was that subsystem goals induced more Fix That Fail decisions.
Goal achievement. The times that subjects achieved assigned goals was used as
performance measurement. The hypothesis was that wholesystem goals were beneficial for
keeping the systems under steady state that is reaching the goals.
Results
The analysis results supported the hypothesis that goal setting made difference on behavior
and performance. Subjects in prey/predator ratio goal condition made more dysfunctional
decisions and thus performed worse than prey/predator number goal condition.
Fix That Fail decisions
As shown in Table 1, subjects made more dysfunctional Fix That Fail decisions in
prey/predator ratio goal condition than in prey/predator number one (X2(2)=4.04, p<0.05).
Subjects accepting prey/predator ratio goal made Fix That Fail decisions because they intended to
decrease prey/predator ratio. While hunting too many prey resulted in the decrease of predator
and thus increased prey/predator ratio consequently as shown in Figure 6. The side effects loop
RI was overlooked although the goal consisted of the number of prey and predator.
Table 1. Fix That Fail decision and goal achievement
prey/predator —_prey/predator prey
ratio goal number goal number Test
setting setting goal setting
Fix That Fail decision 2 28 19 25 X?(2)=4.22,
p=0.12
destruction 4 X2(2)=27.88,
8 ° 5 p<0.005
prey/predator ratio goal F(2,72)=10.4
achievement b il a 0.5 6, p<0.0001
prey/predator number 12 25 1.0 F(2,72)=7.38,
goal achievement > . ” p<0.005
prey number goal F(2,72)=4.14
achievement b a 33 32 1, p<0.05
a: N=41 for each cell; b: Number presented was goal achievement for the final 5 decisions.
= predator +
predator» PUY Predator ___ > hunting
side effect oi
loop R1 policy
: + loop Bl
prey
Figure 5. Fix That Fail induced by the prey/predator ratio goal
711
Given subjects’ bounded rationality of dynamic complexity, a wholesystem goals such as
prey/predator number goal can not stop Fix That Fail decisions thoroughly as shown in Figure 6.
In order to decrease the number of prey, subjects hunted too many prey to decrease predator
unexpectedly and increase prey consequently. More subjects made this kind of dysfunctional
decisions in prey number goal condition than in prey/predator number one.
hi
prey Poop BI prey hunted
fry
side effect pre
y as food
Prey predated loop RI for predator
+ predator
Figure 6. Fix That Fail induced by the prey/predator number and prey number goals
Goal achievement
As shown in Table 1, subjects accepting prey/predator number goals outperformed the
other two groups no matter what goal achievement measurements. The other performance
measure that can be used is to look at the times of system destruction where subjects made too
inferior decisions to destroy the ecosystem. There were eighteen subjects (45%) destroyed the
system in prey/predator condition that was significantly more than the other two conditions.
Discussion and Conclusion
The hypothesis that the choice of goals affects subjects’ decision behavior and performance
‘was supported in the study. Prey/predator number goal, a wholesystem: goal, induced less Fix
That Fail decisions and performed better than the other two types of goals. Although two main
stock variables were included, prey/predator ratio goal setting led subjects to pay attention to prey
or predator subsystems only and make Fix That Fail decisions to destroy the simulated system
frequently. When goals were set with the number of stocks, prey/predator number goal led
subjects to pay attention to both the two main subsystems of prey and predator and decrease Fix
That Fail decisions. Further, the choice of goal just from the subsystem point of view is
insufficient as well. Prey number goals led subjects to make Fix That Fail decisions and perform
worse.
In conclusion, the choice of goals affects decision-makers’ focus of attention. Given
people are bounded rationally on dynamic complexity and goals oriented , an inferior choice of
goal causes decision-maker to fail to manage the systems successfully; some dysfunctional
systems archetype behavior resulted. Choosing the right goals, people could be good dynamic
decision-makers. |
References
Doerner, D. (1980). On the difficulties people have in dealing with complexity. Simulation &
Games, 11, 76-106.
Forrester, J. W. (1995). Counterintuitive behavior of social systems. Working paper D-4468-1
MIT, USA.
Locke, E. A., and Latham, G. P. (1990). A Theory of Goal Setting and Task Performance. New
York: Prentice-Hall.
Sterman, J. D. (1989b). Misperceptions of feedback in Dynamic Decision Making. Organizational i
Behavior and Human Decision Process, 43, 301-335.
Yang, J. (1996). Facilating learning through goal setting in a learning labarotory. Proceedings of
the 1996 International System Dynamics Conference, 593-596.
712
