Parallel Program
Designing the Learning Environment of Learning Laboratories:
Cognitive Strategy, Learning and Transfer
Showing H. Young, Associate Professor
Sy-Feng Wang, Ph.D.
Department of Business Management,
National Sun Yat-Sen University, Kaohsiung, Taiwan
ABSTRACT
The problems of video game syndrome has been an obstacle to prove the value of Manage-
ment Flight Simulators . This paper proposed a theoretical perspective of cognitive strategy to
explain this phenomenon: that is, due to the reasons of (1) rational allocation of limited cognitive
resources, (2) passive generation altemative methods when failed, (3) faulty mental model to
represent the dynamic complexity, the cognitive strategies used by subjects, e.g., feedback con-
trol, feedforward control and memory control, are different from the cognitive strategy of mental
model simulation expected by researchers. Task salience and transfer-oriented task setting were
manipulated to facilitate learning with provoking the appropriate cognitive strategy. The effects
of these two learning aids are tested by one laboratory experiment, and tested by multiple index
with multiple measurement methods. Experimental results support the proposed theoretical
perspective. The mental model simulation strategy seems not the natural cognitive strategy used
by subjects. The learning aids had significant positive effects on inducing the cognitive strategy of
mental model simulation, on the learning of cognitive skill of systems thinking, on the
improvement of task's performance, and on the transfer performance in two transfer tasks.
Introduction
Management Flight Simulators (MFS's) has obtained more and more attentions in the system
dynamics field. However, the popularity of these simulators has far outstripped the research on
their effectiveness. MFS's are effective when they engage people in what Deway called "reflective
thought" and what Sch6n calls "reflective conversation with the situation" (Sterman, 1994).
However, a commonly observed behavior in MFS's is "trial and trial again" or so called "video
game syndrome", where players tend to treat MFS as a video game and rapidly try many different
actions without reflection. They do not take time to reflect on the outcomes, identify discrepan-
cies between the outcomes and their expectations, formulate hypotheses to explain the discrepan-
cies, and then devise experiments to discriminate among the competing alternatives. They simply
keep trying until "their score" improves (Isaacs and Senge, 1992; Sterman, 1994). In such
conditions, how can we feel confident on the effectiveness of MFS?
Since the problem of video game syndrome is so threatening to recognize the effectiveness of
MES, investigating the underlying mechanism of the phenomenon is a very important task before
researching about MFSs' effectiveness. This study aimed at the investigation of the video game
syndrome, particularly focused on the underlying cognitive processes behind the phenomenon,
and on the methods to overcome it. In the following context, we firstly discussed the theoretical
perspectives on the video game: syndrome in the literature, then proposed a new perspective to
see this problem. The video game syndrome would be redefined as a phenomenon of dissociation
between performance and learning in dynamic complexity task. Based on the new definition, a
theoretical explanation was proposed from cognitive point of view. Task salience and transfer-
976
System Dynamics ‘95 — Volume II
oriented task setting were proposed as prescriptive methods to overcome the problem.
Experimental methods were employed to examine the effect of the proposed treatments.
The theoretical perspectives on the video game syndrome
There were three theoretical perspectives about the video game syndrome in the related literature.
Firstly, Isaacs and Senge (1992) argued that the video game syndrome was just the tip of the
iceberg, it was caused by the defensive mechanism existed in individual, group and organiza-tional
level. Participants and interventionists bring Model 1 ways of operating into the settings of MFSs'
learning environment and learning methods. The video game syndrome in MFSs' environment just
reflects the very tendency of human's Model 1 behavior and defensive routine (Argyris, 1990).
Although there were no experimental data to support Isaacs and Senge's argument, we
believed this perspective had explanatory power especially when using MFS in organizational
workshop. However, it was not the whole story. The video game syndrome was also existed in
school's education system and in laboratory experiment (Paich and Sterman, 1993; Wang and
Young, 1992; Young, et al., 1992), where the defensive need are much lower than that of
manager's workshop in corporate. There seems existed some underlying reasons caused by
humanbeing's cognitive process.
The second perspective lied in the humanbeing's poor cognitive ability to represent the
dynamic complexity task of MFSs. Researches of dynamic decision making show that human-
being's faulty mental models of the task environment cause the dysfunctional behaviors and
misperceptions of feedbacks (Diehl, 1992; Dorner, 1980; Kleinmuntz, 1993; Sterman, 1989a, b).
For example, decision makers may have an "open loop" representation of the task, attributing
endogenous behavior of the system to exogenous events (Sterman, 1989a). Even decision maker
might want to "close the loop", but they seem to be incapable of generating appropriate close-
loop models to represent the system, they seem do not know how to do so effectively (Diehl,
1992, p.291-292). Due to lacking suitable representation, subjects were thus tend to “try and try
again" and the video game syndrome occurred.
As we would discussed here, lacking suitable representation was one important cause of the
video game syndrome. But, it was not the whole story. For example, if the video game syndrome
was just only caused by lacking suitable representation, then the only thing left to do was how to
aid subjects to shape suitable representation. In deed, offering systems thinking's tool to aid
subjects to represent the task and to involve into the modeling process, was suggested by many
system dynamists (Graham, et al., 1992; Senge and Sterman, 1992; Vennix, 1990). However,
Wang (1994a) had offering systems thinking's tool to his experimental group, the experiment
results showed that the treatments' effect were covered by the tendency of the cognitive strategy
of feedforward control (see following discussion). There existed similar results in the experiment
of Paich and Sterman (1993), a large part of performance improvement was caused by the
knowledge of last trial's demand pattern, not by the deeper understanding of task's dynamic
structure. The aid of systems thinking's representation might had it's potential effect when
subjects used the cognitive strategy of mental model simulation. When using mental model
simulation, subjects produce a mental model to represent the task system based on their
information and knowledge. Subjects formulate decision policy from the model and test it on the
MEFS, and they can modify their mental model based on decision outcomes (Isaacs and Senge,
1992). If subjects do not use the mental model simulation strategy, e.g., they used trial and error,
977
Parallel Program
how could we expect systems thinking's tool would aid subjects to formulate effective mental
model? There is no represented mental model of the task system in subject's mind.
Therefore, if we could not clarify what cognitive strategies were used by subject and how to
induce the expected mental model simulation strategy, the problem of video game syndrome
seemed could not be completely solved. This paper would propose a third theoretical perspective
about the video game syndrome, which lied in the cognitive strategies used by subjects.
The theoretical perspectives of cognitive strategy
Redefine the problem: the dissociation between performance and learning
From cognitive point of view, the video game syndrome might be redefined as a phenomenon of
dissociation between performance and learning in dynamic complexity task. The phenomenon
demonstrates that practice improved subjects’ performance significantly but had no effect on the
inquiry of task knowledge (Berry and Broadbent, 1988; Sanderson, 1989). Recent experimental
results in the task of MFSs supported such kind of definition. Paich and Sterman (1993) found
that subjects' performance was significantly improved through practice, but little deeper learning
was detected. Wang and Young (1992) had similar findings that performance was dissociated
with task specific knowledge.
When redefined the problem as a phenomenon of dissociation between performance and
learning, we found there existed a few serial researches (e.g., Berry, 1991; Berry and Broadbent,
1984, 1987, 1988) concemed about the dissociation between performance and leaming and the
underlying cognitive processes behind the dissociation phenomenon in simple dynamic control
task . In next section, we will discuss those research results, and then used them to MFS's.
The cognitive process behind the dissociation in simple dynamic control task
A series of studies by Berry and Broadbent (e.g., Berry, 1991; Berry and Broadbent, 1984,
1987, 1988) have suggested the dissociation between task performance and associate verbalizable
knowledge in simple dynamic control task. The typical tasks used by Berry and Broadbent are
combined by a set of linear equations. For example, in the task of Sugar Factory, P = (2W-Pt-1)
+ Random Value (1, 0 or -1) , where P is the production, W is the workforce which only can vary
from 100, 200,...to 1200. Subjects are asked to use W to control P to reach and maintain a target
value. The task knowledge to be learned is the relations of polarity and/or quantity between
decision variables and objective variables.
They showed that practice significantly improved the ability to control the task, but had no
effect on the ability to answer post-task written questions. In contrast, verbal instruction on how
to reach and maintain the target value significantly improved the ability to answer questions but
had no effect on control performance. Moreover, there was an overall significant negative corre-
lation between task performance and question answering. The findings were similar to those
found by some system dynamists, except for the tasks used by system dynamists were more
complicated.
Two possible cognitive processes were adopted to explain the dissociation (for more detail,
see Sanderson, 1989). The first lies in the distinction between explicit and implicif, modes of
learning. That conscious self-report task specific knowledge is not available, because some
information processing is done unconsciously. This is related to the long-standing idea that
cognitive activity takes place in parallel at multiple levels. Another explanation lies in the idea of
978
System Dynamics '95 — Volume II
production-system that verbal knowledge might decay in the process of cognitive skill acquisition.
As learning progresses, simple productions are replaced by more complex, inclusive productions
through the knowledge-compilation process. However, the simple productions can support
verbalizable knowledge about performance, but the more complex one can't, because the latter
compresses a large number of initiating conditions and resulting actions. This explanation is
similar to the idea that human cognitive capacity is limited, thus only the most salient information.
will be processed and reported.
Different cognitive strategies may be a cause to lead to the foregoing two cognitive processes.
Broadbent, et al. (1986) proposed two kinds of cognitive strategies, namely, model manipulation
and situation matching. For the model manipulation strategy, subjects have known relations
among variables. Thus, the strategy can proceed by calculating the future consequences of each
possible action, using the observation of the current situation and the knowledge of the structure
of the world, and then choose the best one. For the situation matching strategy, subject stores a
previously generated table that records the correct action to be taken in each of an array of
situations, Using the current situation as input, subject can lookout a better action from the
"situation-action-performance" table. Model manipulation strategy is based on task knowledge,
then explicit learning occurred. Subjects can modify their task knowledge through the comparison
between forecasts and outcomes. While the understanding about task systems is not necessary for
situation matching, the only thing subject must do is to accumulate the "situation-action-
performance" table, explicit learning thus does not occur. The situation matching process may be
done unconsciously, so that conscious self-report of task specific knowledge is not available or
the productions of situation matching become too complex to support verbalizable knowledge.
In short, the relations between performance's improvement and task knowledge's learning
depend on the cognitive strategy used. The argument is comprehensive to the dissociation in MFS
and will be discussed later.
Cognitive strategies most frequently employed in dynamic complexity tasks
As discussed previously, whether the relations between performance's improvement and task
knowledge's learning are associate or not, depend on the cognitive strategy used. However, due
to the difference between the tasks used by Berry and Broadbent and MFS researches, the
cognitive strategies employed in the dynamic complexity task of MFSs are different.
For the situation matching strategy, since the interdependence and the shift of dominant
loops in MFSs' dynamic complexity task, using the situation matching strategy in MFS's task is
not so effective than used in Berry and Broadbent's task. For example, suppose one subject uses
two cues to identify the situation. Based on the current situation of these two cues, he chooses
one decision numerical value from the stored "situation-action-performance" table. However, this
decision value may bring high score when a certain loop dominated, but worsen performance
when other loop dominated. In deed, the timing of decision is at least as important as the
numerical value of decision in the dynamic complexity task of MFS's.
Feedback control, feedforward control and memory control were found to be used often in
MFS tasks (e.g., Brehmer, 1990; Paich and Sterman, 1993; Sterman, 1989a, b; Wang, 1994a, b;
Wang and Young, 1992). For the feedback control strategy, system structure is treated as a
black box when subjects use feedback control. While using feedback control, no more than the
knowledge of polarity relations between decision and objective variables is needed to approach
the goal. The pattern of decision behavior in the use of feedback control is similar to a "anchoring
979
Parallel Program
and adjustment" under the framework of the goal-seeking negative feedback loop (Kleinmuntz,
1993, p.228). The efficacy of feedback control strategy depends on whether the decision negative
feedback loop can dominate the system.
Feedforward control is similar to feedback control that system structure is treated as a black
box. To use feedforward control, forecasting based on historical data, theory, or expert's experi-
ence is the base to make decision rather than on outcomes in feedback control. For example, the
pattern of production life cycle was used by subjects in Paich and Sterman's study (1993); Books'
law was used by subjects in the study of Sengupta and Abdel-Hamid (1993); forecasting by
experience of previous trial's pattern of system behavior in Wang and Young's study (1992).
These ways of control need lower level of cognitive effort comparing to mental simulation where
understanding about systems structure is necessary (Brehmer, 1990).
For the memory control strategy, subject test some aggregated alternatives by trial and error
and memorize their effects. For example, the pricing decision in one game trial may be aggre-
gated as some alternatives, e.g., low, median and high prices. If the pricifig decision is the
leverage of the system, then performance will be improved by testing, memorizing and selecting
alternatives, but without understanding system's structure(Wang, 1994c).
Although these three cognitive strategies (feedback control, feedforward control and memory
control) are preferred by subjects and advantageous for the improvement of performance, they are
not helpful for deeper learning in MFS.
In contrast, the expected cognitive strategy by system dynamists is mental model simulation
for its theoretical effectiveness for learning (Isaacs and Senge, 1992; Sterman, 1994). When using
mental model simulation, subjects produce a mental model to represent the task system based on
their information and knowledge. Subjects formulate decision policy from the model and test it on
the MEFS, and they can modify their mental model based on decision outcomes (Isaacs and Senge,
1992). Then, the learning about the task system occurs.
To use model manipulation strategy is difficult in MFS. The distinction between model
manipulation and mental model simulation lies in the representation of task where the former
represents task with mathematical type, the latter with a way which is comparable with rule of
human thinking. Subjects can simulate policies for a long-term period with mental simulation but
just one period decisions with model manipulation in MFS because of the complexity of task. In
fact, subjects could hardly use model manipulation in MFS because subjects just can not compute
the high order and nonlinear differential equations in MFS. Therefore, model manipulation is
ignored in the following discussions.
The cognitive causes of the cognitive strategies’ tendency
The foregoing discussions demonstrate that dissociation in MFS resulted from the cognitive
strategies chosen by subjects are not the expected ones by system dynamists. Three main
explanations, but were not mutually exclusive, for subjects tend not to use mental model simula-
tion, had been offered in the literature (Diesel, 1992; Kleinmuntz, 1993) .
(a) people consciously make a cost-benefit trade-off of limited cognitive resources: Although
the cognitive strategy of mental model simulation can obtain higher decision performance than
other strategies, it must spend much more cognitive resources than. others. Subject may rationally
allocated his cognitive resources on the consideration of cost-benefit ratio, and thus choose the
cognitive strategies of memory control, feedback control or feedforward control, but not the
costly strategy of mental model simulation (e.g., Brehmer, 1990) .
980
System Dynamics '95 — Volume II
(b) people passively generate alternative methods, which they test by experiment and
abandon only when failed: The delay times for develop an effective strategy of mental model
simulation, memory control, feedback control and feedforward control were different. To devel-
ope an effective mental model simulation takes much longer time than other three strategies.
Even subjects want to employ the mental model simulation strategy, there are opportunities for
subjects to passively develop another effective strategy (memory control, feedback control or
feedforward control) before the shape up of mental model simulation. It is difficult for human
being to abandon an effective method, thus subjects tend to hold this strategy until it break down
(e.g., Wang, 1994a) .
(c) people rely upon faulty mental models that do not capture the dynamic complexity nature
of the task; Researches of dynamic decision making show that humanbeing's faulty mental models
of the task environment causes the dysfunctional behaviors and misperceptions of feedback
(Brehmer, 1990; DOmer, 1980; Kleinmuntz, 1993; Sterman, 1989a, b). Due to lacking suitable
representation, subjects had difficulty in formulating mental model that could capture the dynamic
complexity nature of the task. Thus subjects were either tend not to use the mental model
simulation strategy, or they want to use it but do not know how to do it effectively (Diesel, 1992).
Methods to overcome the dissociation
Task salience for dynamic complexity task
Rather than simply demonstrating dissociation, an alternative approach has been to look at
conditions that give rise to either implicit or explicit learning. Berry and Broadbent (1988)
propounded that "salience" of task could affect the used cognitive strategy. They found low
salience led to implicit learning, and the relation between performance and task knowledge is
vague or even negative, vice versa. Task salience, defined by Berry and Broadbent, is the
probability that, if a person learns by the explicit rather than the implicit mode, the key variable
will be chosen. There are three ways to increase the level of task salience as follow:
(1) To reduce irrelevant factors in situation (Broadbent, et al., 1986); for example, to reduce
the number of relations of variables to be processed in a decision.
(2) To make the key events act in accordance with general knowledge from outside the task;
for example, to remove the delay between actions and outcomes (Berry, 1991; Berry and
Broadbent, 1988), or to add a positive feedback loop to increase the impact of actions on
outcomes (Broadbent, et al., 1986).
(3) To give an explicit verbal direction as to which are the key variables; for example, to
instruct subjects what kind of variables are relevant (Berry and Broadbent, 1988).
Accordingly, for system dynamists, it is possible to lead subjects to use the expected cognitive
strategy through the manipulation of task salience in order to overcome the dissociation in MFS.
Nevertheless, the manipulation of task salience should be modified, because task properties in
MFS are different from those in the research by Berry and Broadbent.
The manipulation of task salience to induce mental model simulation strategy in MFS. is
possible. First, to provide subjects reference mode of the task system can increase task salience,
because the key variables and their pattern of behavior are given. Second, to provide causal loop
diagram can eliminate redundant information and hint subjects the polarity relations and delay
between those key variables. Furthermore, causal loop diagram can instruct subjects how to
981
Parallel Program
represent a complex dynamic system, and reduce the barrier of using mental model simulation.
Third, partial model test divides a whole complex system into several controllable parts and thus
increase the salience of task (for more detail, see Young, et al., 1994). This design is similar to
that in the study by Broadbent, et al. (1986) where subjects were instructed to test the relations
between variables once at a time.
Attitude toward learning and transfer-oriented task setting
Attitude toward learning is the other factor to affect the use of mental model simulation
strategy (Isaacs and Senge, 1992; Kleinmuntz, 1993). Subjects who are willing to learn will more
likely put more cognitive resources to use mental model simulation. In contrast, sibjects who
have lower motivation to learn may apt to use memory control, feedback control and feedforward
control to avoid severe exertion.
Goal setting, as suggested by Kleinmuntz (1993) and Brehmer (1992, p.238), may be one
method to motivate subjects’ motivation to learn. However, in Wang's (1994c) experiment,
although high motivation was induced by goal setting, but the induced motivation and cognitive
effort might be used in the wrong place. Some subjects spent their time in the cognitive strategy
of memory control, but not the expected mental model simulation. They found some way to get
high score, although they do not know why. The setting goal was satisfied, but the learning was
not occurred. In the present study, "transfer-oriented task setting" is used to replace the method
of goal setting.
Experiment Design
Table 1 describes the experimental design. The experiment had two manipulated variables and
one block variables, thus shape a 2*2*2 proportional full factorial between-subjects design. The
manipulation of fask salience was contrast by non-salience. The manipulation of transfer-
oriented task setting was contrast by control-oriented task setting. Subjects of 24 MBA students
were randomly assigned to four cells of these two manipulated variables. Subjects of 20
undergraduate students, majoring in business management, were also randomly assigned to those
four cells.
Table 1. Experimental Design
MBA students undergraduate students
transfer-oriented__control-oriented transfer-oriented__control-oriented
task salience 6 6 5 5
non-salience 6 6 5 5
STRATAGM-2 was used to run the experiment(for details, see Sterman, 1989b, Sterman and
Meadows, 1985). All the information of 13 variables of STRATAGM-2's model were shown in
computer's monitors. Subjects made decision on computer. The 44 subjects were paid volunteers
from National Sun Yat-Sen University in Taiwan, aged between 21 and 28. None had
participated in such experiments used dynamic complexity task.
Subjects were asked to finish 4 trials (25 periods in one trial) of the STRATAGEM-2 :task.
The manipulations were manipulated during these four trials. After these four trials, one
questionnaire and two transfer tasks (play 2 trials each task) were given to evaluate the used
cognitive strategies and the learning performances.
982
System Dynamics '95 — Volume II
Manipulation
Subjects in the group of task salience received an animated pattern-structure explanation in
the end of every trial, while subjects in the contrast group’ of non-salience do not received this
material. The explanation included the typical behavior when subjects first time interact with the
task (see Sterman, 1989b), and the causal loops that produced those pattern of behavior.
Subjects in the ransfer-oriented group received the task setting that "your task is to learn to
transfer into the transfer task, the rewards are paid based on the decision performance of the
transfer task." Subjects in the control-oriented group were not announced about the transfer task,
and the every trial's decision performance of the task will influence their reward. They are forced
to calculate their reward in the end of each trial.
Dependent Variables
As shown in Table 2, there are four dependent variables in the experiment: the decision
performance, the learning of systems thinking, the tendency of using mental model simulation
strategy and the leaming transfer of the changing-goal loop. The dependent variables were tested
by multiple index with multiple measurement methods. For example, the /earning of systems
thinking is constructed by whether or not subject can perceive and/or treat the four dominance
loops (as shown in Table 2), which are assumed to be the basic cognitive schema subjects might
learn and can transfer to other situations (e.g., the learning transfer of the changing-goal loop in
Table 2). The perception and treatment of almost every one loop are operationalized by three
measurement methods, include the decision rule analysis, protocol analysis of the cognitive map,
and the scenario testing index. For the limitation of pages, we will only introduce the methods of
the decision rule analysis (for the entire measurement method, see: Wang, 1994b).
As shown in Table 3, there are three groups of decision policies in the task. The first is the
typical behavior as observed by Sterman (1989b) that increase capital to satisfy demand. The
dominance loops are the changing-goal loop A and, if considered, the supplyline-adjustment
loop. Based on the work of Sterman (1989b), the perception and treatment about these two
loops can be observed by the variance conditions (across trials) of two parameters s1 and p.
The second group of decision policies is to increase demand to suit surplus capital. As shown
in Table 3, based on different cue, there are three policies to do this. Due to the basis for
comparison are different, the parameters of subjects decision rules are not suitable to represent
the perception and treatment about the changing-goal loop B. Since this group of policies are
exclusive with the first one, we decide to use the method that "What is the cost due to the use of
these types of policies instead of the first one." Figure 1 showes the method... ‘If subject.keeps to
use the origin policy, the performance index is 1559. However, this subject use another policy
after period 15, and obtain the performance index as 2662 . So, the cost of this policy (due to he
does not perceive and/or treat the changing-goal loop B) is 1103 (2662-1559).
The third group of decision policies is tend not to react to change, for example, keep constant
decision value. All these three groups of policies are exclusive to each other, but subjects can use
them in the different periods in one trial. Thus, we also use the cost index to treat this group of
policies.
Results and Conclusions
Experimental results showed that these two learning aids had significant positive effects on
983
Parallel Program
Table 2. The measurement methods of the dependent variables
Dependent
Variables Index Sub-Index
decision performance
perception and treatment about
* the changing-goal loop A *parameter s1 (from decision rule analysis)
*the report of the changing-goal loop A
*the changing-goal loop B *cost index 1 (from decision rule analysis)
the learning of *the report of the changing-goal loop B
systems thinking *scenario test index: capital surplus
*the implicit loop *cost index 2 (from decision rule analysis)
*the report of the implicit loop
*scenario test index: second wave
*the supplyline-adjustment loop “parameter p/s] (from decision rule analysis)
*the report of the treatment level of delay
*scenario test index: supply will over demand
*the block box memory control
*the quality of the policy theory *number of concepts *number of task's concepts
in the cognitive map *number of new concepts *number of relationships
the tendency of *% of polarity relationships *% of correct relationships
jusing mental * number of material relationships *number of information relationships
Imodel *number of material paths *“number of information paths,
simulation *length of material paths *Jength of information paths
*the existence of the control loop
*cognitive resource's allocation _*time taken per trial
*the active attitude to try *interaction between trials and adopt policy
divergent policy *trials with planning
the leaming _ *the learning transfer in the *decision value’s mean, standard deviation, maximum, average fluctuation|
transfer ofthe _ transfer task A *decision performance
changing-goal *the learning transfer in the *decision value's mean, standard deviation, maximum, average fluctuation]
loop transfer task B *policy of no-action
ise oo Fe
Table 3. The decision policies 35 10° unit policy of group 2,
Policy Group Decision policy used by subject #33
[group 1 ID=CD+s1(BT-KI}+p(CD*DD-BK)
group 2 ID=s2a*(KI-BT) 20
(only existed when KI>BT) |D=s2b*(KI-BG) 7 policy of group 1,
[D=s3c*(KI-BT+2CA-CD) 154 original used in
group 3 (D=costant 1~15 periods
(only existed when KI<BT) |D=costantts3*(KE-BT) “10 wae
Note: BG: Backlog of Good sector CD: Capital Depreciation capital inventor
BK: Backlog of Capital sector D: Decision 5 | backlog-total
BT: Backlog-Total DD: Delivery Delay : A
CA: Capital Acquisition KI: Capital Inventory nd
10 15 20 25
Period
Figure 1. Measurement Method of Cost Index
inducing the cognitive strategy of mental model simulation, on the learning of cognitive skill of
systems thinking, on the improvement of task's performance, and on the transfer performance in
two transfer tasks. Based on those experimental results, it seemed safely to conclude that:
without any learning aids, the "natural" cognitive strategy subjects used in dynamic complexity
task was not the mental model simulation strategy. In contrast, the cognitive strategies of
memory control, feedback control and feedforward control seem were mostly employed by
subjects. Researchers of MFS's should not expect that subjects would use the mental model
984
System Dynamics '95 — Volume II
simulation strategy automatically, we should take the theoretical perspective of cognitive strategy
into consideration .
Reference
Argyris, C. 1990. Overcoming Organizational Defenses. Boston, MA: Allyn and Bacon.
Berry, D. C., and D. E. Broadbent. 1988. Interactive Tasks and the Implicit-Explicit Distinction. British Journal of Psychology
79: 251-272.
Berry, D. C. 1991. The Role of Action in Implicit Learning. Quarterly Journal of Experimental Psychology 43A (4): 881-906.
Berry, D. C., and D. E. Broadbent. 1984. On the Relationship between Task Performance and Associated Verbalizable
Knowledge. Quarterly Journal of Experimental Psychology 36A: 209-231.
Berry, D. C., and D. E. Broadbent. 1987. The Combination of Implicit and Explicit Learning Processes. Psychological
Research 49: 7-15.
Brehmer, B. 1990. Strategies in Real Time Dynamic Decision Making. in R. M. Hogarth ed. Insights in Decision Making,
262-279, Chicago: The University of Chicago Press.
Brehmer, B. 1992. Dynamic Decision Making: Human Control of Complex Systems. Acta Psychologica 81: 211-241.
Broadbent, D., P. FitzGerald, and M. Broadbent. 1986. Implicit and Explicit Knowledge in the Control of Complex Systems.
British Journal of Psychology. 77: 33-50.
Démer, D. 1980. On the Difficulties People Have in Dealing with Complexity. Simulation & Games 11: 76-106.
Diesel, E. 1992. Effects of Feedbacks Structure on Dynamic Decision Game. Ph.D, Dissertation, Sloan School of Management,
MIT, Cambridge, MA, USA.
Graham, A. K., J. D. W. Morecroft, P. M. Senge, and J. D. Sterman. 1992. Model-Supported Case Studies for Management
Education. European Journal of Operational Research 59 (1): 151-166.
Isaacs, W., and P. M. Senge. 1992. Overcoming Limits to Learning in Computer-Based Learning Environment. European
Journal of Operations Research 59 (1): 183-196.
Kleinmuntz, D. N. 1993. Information Processing and Misperception of the Implications of Feedback in Dynamic Decision
Making. System Dynamics Review 9 (3); 223-238.
Paich, M., and J. D. Sterman. 1993. Boom, bust, and failures to learn in experimental market. Management Science 39(12):
1439-1458.
Sanderson, P. M. 1989. Verbalizable Knowledge and Skilled Task Performance; Association, Dissociation, and Mental Model.
Journal of Experimental Psychology: Learning Memory and Cognition. 15: 729-747.
Senge, P. M., and J. D, Sterman. 1992. Systems Thinking and Organizational Leaming: Acting Locally and Thinking Globally
in the Organization of the Future. European Journal of Operations Research 59 (1): 137-150.
Sterman, J. D. 1989a, Modeling managerial behavior: Misperceptions of feedback in a dynamic decision making experiment.
Management Science 35 (3): 321-339.
Sterman, J. D. 1989b. Misperception of feedback in dynamic decision making. Organizational Behavior and Human Decision
Process 43: 301-335.
Sterman, J. D. 1994. Leaming in and about Complex Systems. System Dynamics Review 10 (2): 291-330.
Sterman, J. D., and D. L. Meadows. 1985. STRATAGM-2: A Microcomputer Simulation Game of the Kondrative Cycle.
Simulation and Games 16 (2): 74-202.
Vennix, A. M. J. 1990, Mental Models and Computer Models: Design and Evaluation of a Computer-Based Learning
Environment for Policy-Making, Ph.D. Dissertation, the Nijmegen Institute for Cognition Research and Information
Technology, University of Nijmegan, Netherlands.
Wang, S. 1994a. A Prescriptive Experiment in Dynamic Complexity Task: Learning Laboratory, Systems Thinking, and
Dissociation between Performance and Learning. Stn Yat-Sen Management Review. 2 (3): 97-122.
Wang, S. 1994b. Designing the Management Learning Laboratory: Cognitive Strategy, Systems Thinking and Learning
Transfer. Ph.D. Dissertation, the Institute of Business Management, National Sun Yat-Sen University, Kaohsiung, Taiwan.
Wang, S. 1994c. Designing the Management Leaming Laboratory: Overcoming the Dissociation between Performance and
Leaming. Sun Yat-Sen Management Review. 2 (4).
Wang S., and S. H. Young, 1992. A Preliminary Experiment on Examining Thinking in a Meta-Dynamic Decision Making
Environment. in The Proceedings of the 1992 System Dynamics Conference, Utrecht, Netherlands, 757-766.
Young, S. H., J. Yang and S. Wang. 1992. Enhancing the Learning Effects of Dynamic Decision Game on Systems Thirtking:
An Experimental Study. The Proceedings of the 1992 International System Dynamics Conference, Utrecht, Netherlands,
847-856.
Young, S. H., S. Wang, and J. Yang. 1994. Overcoming the Leaming Barriers of Management Flight Simulators. Proceedings
of the 1994 International Conference of the System Dynamics Society: Microworld, Sterling, Scotland, 101-108 .
985
CC BY-NC-SA 4.0