Está en la página 1de 13

Liu, T.-C. (2010).

Developing Simulation-based Computer Assisted Learning to Correct Students' Statistical Misconceptions


based on Cognitive Conflict Theory, using "Correlation" as an Example. Educational Technology & Society, 13 (2), 180192.

180
ISSN 1436-4522 (online) and 1176-3647 (print). International Forumof Educational Technology & Society (IFETS). Theauthors and theforumjointly retain the
copyright of thearticles. Permission to makedigital or hard copies of part or all of this work for personal or classroomuseis granted without feeprovided that copies
arenot madeor distributed for profit or commercial advantageand that copies bear thefull citation on thefirst page. Copyrights for components of this work owned by
others than IFETS must behonoured. Abstracting with credit is permitted. To copy otherwise, to republish, to post on servers, or to redistributeto lists, requires prior
specific permission and/or afee. Request permissions fromtheeditors at kinshuk@ieee.org.
Developing Simulation-based Computer Assisted Learning to Correct
Students Statistical Misconceptions based on Cognitive Conflict Theory, using
Correlation as an Example

Tzu-Chien Liu
Graduate Institute of Learning & Instruction and Center for Teacher Education, National Central University, Taiwan
// ltc@cc.ncu.edu.tw

ABSTRACT
Understanding and applying statistical concepts is essential in modern life. However, common statistical
misconceptions limit the ability of students to understand statistical concepts. Although simulation-based
computer assisted learning (CAL) is promising for use in students learning statistics, substantial improvement is
still needed. For example, few simulation-based CALs have been developed to address statistical
misconceptions, most of the studies about simulation-based CAL for statistics learning lacked theoretical
backgrounds, and design principles for enhancing the effectiveness of dynamically linked multiple
representations (DLMRs), which is the main mechanism of simulation-based CAL, are needed. Therefore, this
work develops a simulation-based CAL prototype, Simulation Assisted Learning Statistics (SALS), to correct
misconceptions about the statistical concept of correlation. The proposed SALS has two novel elements. One is
the use of the design principles based on cognitive load and the other is application of the learning model based
on cognitive conflict theory. Further, a formative evaluation is conducted by using a case study to explore the
effects and limitations of SALS. Evaluation results indicate that despite the need for further improvement, SALS
is effective for correcting statistical misconceptions. Finally, recommendations for future research are proposed.

Keywords
Simulation-based CAL, Misconception, Cognitive conflict theory, Learning model, Cognitive load, Dynamically
linked multiple representations

Introduction

Understanding and applying statistical concepts is now essential for citizens in information-intensive societies (Gal,
2002; Garfield & Ben-Zvi, 2007; Shaughnessy, 2007). In accordance with this trend, statistics is considered an
important learning topic in most educational levels in many countries (Shaughnessy, 2007).

However, helping students to develop statistical literacy is difficult (Gal, 2002; Garfield & Ben-Zvi, 2007).
Statistical misconceptions are common in students at all levels from elementary to graduate as well as in adults and
even researchers. These misconceptions conflict with scientifically accepted statistical concepts and seriously hinder
the comprehension and application of statistics (Castro Sotos et al., 2007; Cohen, Smith, Chechile, Burns, & Tsai,
1996; Garfield & Ben-Zvi, 2007; Liu, Lin, & Tsai, 2009; Morris, 2001; Shaughnessy, 2007).

Statistical misconceptions are systematic patterns of error in interpreting, understanding or applying statistical
concepts, which may result from language, daily experience, existing knowledge and learning materials (Castro
Sotos et al., 2007; Cohen et al., 1996; Liu et al., 2009). The literature reveals three characteristics of statistical
misconceptions (Cohen et al., 1996; Cumming & Thomason, 1995; Garfield & Ben-Zvi, 2007; Liu et al., 2009;
Morris, 2001): (1) they are difficult to detect; (2) they are difficult to correct; and (3) they impede further learning of
statistics. Therefore, statistical misconceptions are extremely challenging to educators who seek to enhance the
statistical literacy of their students. Given these considerations, helping students to redress their statistical
misconceptions is an important research issue.

Recent studies have proposed simulation-based computer assisted learning (CAL) for helping students learn statistics
(e.g. Cumming & Thomason, 1995; Morris, 2001). Simulation-based CAL is a learning environment that combines
learning guides and Dynamically Linked Multiple Representations (DLMRs). Using DLMRs to learn statistical
concepts (e.g., correlation) enables students to interact with one representation (e.g., changing the value of a
correlation coefficient) and receiving instant feedback from other representations (e.g., the corresponding change in a
scatter plot and a table with x and y value). Scholars suggest that such linking of abstract ideas (e.g., correlation
coefficient) with concrete representations (e.g., scatter plot) can help students understand statistical concepts (e.g.,
correlation) (Meletiou-Mavrotheris, 2004; Mills, 2002).
181
Despite the recognized potential of simulation-based CAL for helping students to understand statistics, the design
and development of simulation-based CAL for learning statistics still warrant further study (Kadijevich, Kokol-
Voljc, & Lavicza, 2008; Mills, 2002; Morris, 2001; Morris et al., 2002). For instance, Shaughnessy (2007) and
Castro Sotos et al. (2007) concluded in literature reviews that holding statistical misconceptions is a major
obstruction to comprehending statistics. However, only few simulation-based CALs (e.g., Cumming & Thomason,
1995; Morris, 2001) have been designed to correct statistical misconceptions.

Besides, an effective learning technology requires learning models based on appropriate theory. However, earlier
studies (e.g., Mills, 2002; Morris et al., 2002) have noted the lack of theoretical background for designing and
developing simulation-based CAL to enhance statistical understanding.

Finally, DLMRs is a core mechanism of simulation-based CAL. Applications of DLMRs for learning in other fields
reveal limitations such as passive learning (e.g., Ainsworth, 1999) and cognitive overload (e.g., Lowe, 1999).
Therefore, the design of simulation-based CALs for correcting statistical misconceptions should maximize the
advantages of DLMRs and minimize its disadvantages.

For the above reasons, this study had three major objectives. First, a literature review was performed to establish the
design bases of simulation-based CAL for correcting statistical misconceptions, including the background theory for
learning model and the design principles for simulation-based learning environment. Moreover, a simulation-based
CAL prototype, namely, Simulation Assisted Learning Statistics (SALS), was designed and developed. Major
features of SALS include (1) designs and functions for enhancing the effectiveness of DLMRs, (2) a learning model
based on cognitive conflict approach for correcting statistical misconceptions, and (3) ten learning activities based on
the learning model and earlier studies about statistical misconceptions. Finally, a formative evaluation was conducted
by a case study to explore the effects and limitations of SALS. The formative evaluation was intended to reveal
possible improvements in the proposed SALS as well as additional topics for future research.

When developing SALS, correlation was selected as the focal statistical concept for four reasons. First, correlation is
an essential statistical concept. Second, some misconceptions about correlation are common among students at many
educational levels (Liu & Lin, in press; Liu et al., 2009; Mevarech & Kramarsky, 1997; Morris, 2001; Shaughnessy,
2007). Third, correlation is suitable for learning with DLMRs because it is often explained using multiple
representations, including graphs, texts, numerical values or symbols. Finally, earlier series studies by the author
explored correlation misconceptions and developed diagnostic tools (e.g., Liu & Lin, in press; Liu et al., 2009).
Correlation is thus a realistic and practical concept for designing SALS.

The next section first presents the theoretical background and design principles of SALS and then introduces its
learning environment, learning model and learning activities. A case study with think-aloud and interview method is
then presented to demonstrate students learning processes in learning with SALS and to survey their perceptions
about SALS. Finally, conclusions are given, and future works are recommended.


Design basis of SALS

Theoretical background for learning model

Conceptual change is essential for meaningful learning (Mayer, 2002), and cognitive conflict is considered a
premise for conceptual change (Hewson & Hewson, 1984; Strike & Posner, 1985). Cognitive conflict occurs when
an individual can not apply their existing concept to solve a problem, and is thus confronted with a situation that
motivates the learning of new concept (Lee & Kwon, 2001). Therefore, cognitive conflict strategies are used in many
practical applications to achieve conceptual change (e.g., Chan, Burtis, & Bereiter, 1997; Limon, 2001). According
to the literature, applications of cognitive conflict strategies for achieving conceptual change could be characterized
by the following four key elements (Chan et al., 1997; Limon, 2001; Strike & Posner, 1985): making students aware
of their existing concepts before instructional intervention, confronting them with contradictory information, using
anomalous data or discrepant events to replace prior concepts with scientifically accepted ones, and measuring the
resulting conceptual change. In the current study, the SALS learning model included four phases: Externalization,
Reflection, Construction, and Application. The application of these four elements in SALS is described in detail
below in the section entitled learning model.
182
Design principles for simulation-based learning environment

The main mechanism of simulation-based CAL is DLMRs, which has potential benefits for statistics instruction
(Cumming & Thomason, 1995; Morris, 2001). Despite the advantages of DLMRs, however, scholars in other fields
have noted major disadvantages of DLMRs, which include passive learning and cognitive overload (Ainsworth,
1999; 2006; Lowe, 1999). This section describes how the simulation learning environment was designed based on
cognitive load-related literature (e.g., Sweller, van Merrienboer, & Paas, 1998; Wouters, Paas, & van Merrienboer,
2008) regarding inciting germane cognitive load (the load that can enhance students to engage in learning),
decreasing intrinsic cognitive load (the difficulty or complexity of the learning material itself) and preventing
extraneous cognitive load (the unnecessary load that hinders learning).

Inciting germane cognitive load by expectancy-driven methods: When learning with DLMRs, students manipulate
one representation, and corresponding representations are changed automatically and immediately. Learning by
this overly automated method may be excessively passive and may limit the ability of students to reflect on
relationships between representations, which then results in inability to construct the required understanding
(Ainsworth, 1999; 2006). Germane cognitive load is the load for facilitating student engagement in learning. A
literature review of Wouters et al. (2008) found that using expectancy-driven methods (such as presenting a question
to the students before their learning) is effective for inciting germane cognitive load. Therefore, the current study
applied expectancy-driven methods to promote active learning with DLMRs.

Decreasing intrinsic cognitive load by appropriately structuring the presentation of learning content: Cognitive
overload is a common problem in learning with DLMRs (Lowe, 1999). Cognitive overload is caused by a cognitive
task that exceeds the capacity of working memory (Sweller et al., 1998; Wouters et al., 2008). When learning with
DLMRs, students must observe different representations, pay attention to the relate changes that occur
simultaneously in the representations, and identify the relationships between different representations based on their
manipulation results. Mayer (2005) and Wouters et al. (2008) pointed out that segmenting the learning content or
activity in an appropriately structured way could reduce intrinsic cognitive load when processing complex learning
tasks. Therefore, a simulation learning environment should be designed to provide explicit hints in a highly
structured manner and to guide students in manipulating DLMRs and observing progressively complex relevant
relations within each representation.

Applying contiguity principle to reduce extraneous cognitive load: Extraneous cognitive load is another cause of
cognitive overload in DLMRs learning. When learning with DLMRs, students must observe and relate changes that
occur simultaneously in different representations, which may result in cognitive overload (Lowe, 1999). Therefore,
reducing extraneous load imposed by inappropriate instructional design and freeing cognitive resources for learning
are key design objectives. Applying the contiguity principle that representations place near corresponding parts of
other representations to reduce visual scanning is viewed as an effective means of reducing extraneous cognitive
load, which should be of priority concern when designing simulation-based CAL (Mayer & Moreno, 2003).


Introduction to SALS

The simulation-based learning environment

The SALS was written in Flash Action Script. The major learning environment of SALS (used in Reflection and
Construction phase) was the simulation-based learning environment (Fig. 1), which included the DLMRs area on the
right and learning guide area on the left. The major purpose of the simulation-based learning environment was to
support hands-on exploration activities in controlled settings. The components in DLMR area included the data
points of the scatter plot, sample size (N), correlation coefficient (r), and the values of variables x and y in the table.
Students could manipulate each component and observe the corresponding changes in other components. For
example, after adding, deleting, moving, or re-sampling the data points, students could observe changes in sample
size, correlation coefficient, or the values of variable x and y.

183

Figure 1: SALS learning environment


Figure 2: Applying design principles of SALS

To enhance the effectiveness of the DLMRs, the simulation-based learning environment reflected the three design
principles (Fig. 2). First, to promote active learning with DLMRs, learning guide was provided in problem situations
to prompt students to perform exploratory activities in which they were encouraged to find answers by manipulating
DLMRs (Fig. 2, point A). To decrease intrinsic cognitive load, the learning guide area provided a highly structured
184
series of prompts to guide students to manipulate and observe specific representations and to lead them to
progressively explore the relationships among the representations (Fig. 2, point B).

Finally, to reduce extraneous cognitive load, the DLMRs in SALS contained various special functions. For example,
the multiple windows function enabled students to display four windows simultaneously in the DLMR area (Fig. 2,
point C), to systematically manipulate the representations in all four windows, and to compare the results. For
another example, passing the cursor over a data point automatically displayed the values for x and y (point D in Fig.
2). Therefore, students could easily determine the values of x and y for each data point without consulting a table.
Because these functions did not require students to keep the intermediate results in their working memory, their
extraneous cognitive load was reduced, which enabled them to focus on the relationships among different
representations.


Learning model

The cognitive conflict approach was used to develop a learning model with four phases (Externalization, Reflection,
Construction, and Application) (Fig. 3). The DLMRs was the main learning tool in the Reflection and Construction
phases and in part of the Application phase. The aim of each phase and the progression between each phase are
described below.
Figure 3: The SALS learning model

Externalization phase: Making students aware of their preconceptions before instructional intervention is the first
step of the cognitive conflict approach. Thus, the objective of this phase is to enable students be aware of their own
ideas about the statistics concepts they are learning. This phase achieves this by posing a question in the context of
daily life so as to make it meaningful to the students. The question is also presented as a role-play: an animated
anchorwoman reports background information related to the question before interviewing the students (Fig. 4). The
students then play the role of interviewees and answer the question (Fig. 5). By involving the students in the situation
underlying the question and by helping them to externalize their views, the role-play motivates students to learn the
target concepts in the following phases.


Figure 4: Externalization phase: an animation is used to give background information before presenting the question.

185

Figure 5: Externalization phase: The students answer the question

Reflection phase: This phase prompts students to reflect on their own ideas that were externalized in the
Externalization phase. To achieve this, the simulation-based learning environment displays a series of questions
related to the concepts raised in the Externalization phase in order to guide them to manipulate the DLMRs and to
observe the results progressively. After completing this task, the learning guide then prompts students to compare
their ideas that were externalized in the Externalization phase with the exploration results obtained in this phase and
to reflect on their differences.

Construction phase: This phase introduces exploratory activities to help students construct their concepts about
correlation. In this phase, clear learning guides (including definitions of the target concept and clear manipulating
procedures) are used to prompt students to manipulate the DLMRs, to observe and identify the relationships among
different representations and, then, to construct their own concepts about correlation. Students who do not
understand the concept at the end of this process are asked to repeat the entire Construction phase.

Application phase: This phase provided students with two problem situations in different contexts and with different
solution paths. When solving the problems, the students were given minimal support. One objective was to allow
students to elaborate on their newly constructed concepts by applying them to solve novel problems. The second was
to evaluate the ability of the students to transfer these concepts. One problem was presented as a hands-on activity
that required the students to solve a problem by applying their concepts to manipulate DLMRs. Another problem was
presented as a multiple-choice item to test the same correlation concept raised by the question in the Externalization
phase, but in a different context. Students who correctly solved both problems were assumed to have understood the
correlation concept and could select the next activity to learn another one. Otherwise, they were guided back to the
Construction phase to relearn the concepts.


Learning activities

Ten learning activities were designed for SALS based on the above learning model and earlier studies about
correlation misconceptions. Each learning activity was associated with a learning topic about correlation and was
designed to correct a corresponding misconception (Table 1). These misconceptions were selected for the following
reasons. First, holding these misconceptions could obstruct their learning of concepts related to correlation. Second,
the misconceptions have been confirmed by earlier studies (e.g., Liu & Lin, in press; Liu et al., 2009; Morris, 2001).
Third, these misconceptions are widely held by students (Liu & Lin, in press). Fourth, as reported by Yu, Behrens
and Anthony (1995), these misconceptions should be easily represented by multiple representations. Fifth, earlier
studies (Liu et al., 2009; Liu & Lin, in press) have suggested possible reasons why students hold these
misconceptions, and these reasons should be considered when designing SALS learning activities.



186
Table 1: Learning topics and corresponding misconceptions related to each learning activity
Activity Learning topics Misconceptions
1 Positive and negative
correlation
A correlation between x and y is positive means that as the value of x
increases (decreases), the value of y certainly increases (decreases).
A correlation between x and y is negative means that as the value of x
increases (decreases), the value of y certainly decreases (increases).
2 Positive and negative
correlation
A correlation between x and y is positive means that if the value of x is
very big, then the value of y should also be very big.
A correlation between x and y is negative means that if the value of x is
very big, and the value of y should be very small.
3 Perfect positive and
negative correlation
As long as the sampling points fall in a straight line (including a
horizontal or vertical straight line), the correlation among the variables is
perfect.
4 Zero correlation A zero correlation between two variables indicates No association
between two variables.
5 Correlation degree A positive correlation is stronger than a negative one.
6 Correlation degree The denser the data points in a scatter plot, the stronger the correlation.
7 Definition of the
correlation coefficient
Correlation coefficient (r) is related to slope (m).
8 Calculation of the
correlation coefficient
The correlation coefficient changes when adding a constant to (or
subtracting a constant from) each value of a variable.
9 Calculation of correlation
coefficient
The correlation coefficient changes if the unit of measurement changes.
10 Calculation of correlation
coefficient
The correlation coefficient changes if the variables designated as x and y
are reversed.


Formative evaluation

Methodology

To examine the effectiveness and limitations of using SALS to correct statistical misconceptions, a formative
evaluation was conducted using a case study approach (Yin, 1994) which is particularly appropriate for exploring
students learning processes in learning with technology (Liu, 2007; Tsui & Treagust, 2003). In the current study, in
addition to comparing student misconceptions before and after SALS learning, data about learning processes
associated with SALS learning were collected by think-aloud method (van Somersen, Barnard, & Sandberg, 1994).
Interviews were then performed to assess their perceptions about using the system. The think-aloud method is
considered useful for collecting data about learning processes and particularly for assessing processes related to
cognition (Davison, Vogel, & Coffman, 1997).


Participants

The short form of the Two-tier Diagnostic Instrument about Correlation Misconceptions (TDICM) was administered
to forty-five undergraduate students enrolled in an Education Psychology class in a public university in Taiwan. The
short form of TDICM has ten items which are selected from TDICM developed by Liu and Lin (in press) to diagnose
ten misconceptions that SALS is designed to correct (see Table 1 for details). The test results showed that the
average number of misconceptions was about five. For exploring the effects of SALS on correcting statistical
misconceptions, the two students who had the most misconceptions about correlation and who agreed to participate
in the study (student A with nine misconceptions and student B with eight) were selected as the study cases. Both
students were female. Student A was a sophomore English language and literature major, and student B was a junior
Chinese language and literature major. The interviews conducted before the formative evaluation indicated that,
despite learning basic correlation concepts in senior high school, both had difficulty in recalling and applying these
concepts.

187
Procedure

The two students were individually instructed to perform the think-aloud method (average training time, 32 minutes)
and to manipulate SALS (average training time, 25 minutes). The two students then used SALS to learn correlation
concepts individually and without a time limit (average time to complete each of the ten learning activities, 8.4
minutes). The students used the think-aloud method to present their learning processes while working with SALS,
and the researchers took notes and video recorded the utterances and manipulation behaviors of the students. After
completing the learning activities, the students immediately took the TDICM (post-test). The researchers then
showed the students video recordings of their learning processes and interviewed them to determine the reasons for
their specific behaviors and expressions. The researchers then interviewed the students to assess their perceptions of
SALS. Finally, all data, including manipulation behaviors, the results of think-aloud, and the interviews were
transcribed into digital protocols for triangulation to enhance validity.


Evaluation results and discussion

The current study utilizes a case study to formatively evaluate whether SALS can correct statistical misconceptions
and how such a system can be optimized. A comparison of pre-test and post-test scores on the TDICM showed that
SALS learning could reduce misconceptions about correlation concepts in both students. All nine misconceptions
held by Student A were corrected while six of the eight held by Student B were corrected. To further explore how
SALS benefited the students and the possible shortcomings of SALS, the following sections display the results and
discussions about the effects of learning model, the contributions of design principles of simulation-based learning
environments, and limitations of SALS.


The effects of learning model based on cognitive conflict approach

The learning model including four phases (Externalization, Reflection, Construction, and Application) was
developed in this study according to cognitive conflict approach. Each phase was designed to achieve specific aim to
make students have successful cognitive conflict for conceptual change. This section describes the learning processes
observed in the fifth learning activity and the subsequent interview results to show how the learning model can
benefit the students. The fifth activity is designed for addressing the misconception that a positive correlation is
stronger than a negative one, which is among the most common misconceptions (e.g., Liu & Lin, in press; Liu et al.,
2009; Morris, 2001). For brevity, the following discussion focuses on student A in the last two phases (Construction
and Application) since the two students demonstrated similar learning processes in these two phases of the fifth
activity.

Externalization phase: The SALS used role-play (i.e., interview by the animated anchor) in this phase to present a
question in the context of the daily life experiences of the students to make them to externalize their existing ideas
and to motivate them to continue exploring possible answers by manipulating DLMRs in next two phases. For
example, in the Externalization phase of the fifth learning activity, the animated news anchor described a study of the
relationship between two variables, rate of attendance and statistical achievement. This broadcast displayed four
correlation coefficients (r=-0.9, -0.5, 0, and 0.6) for the two variables in four different high schools. The students
were then asked to select one of four options to indicate the strength of the four correlation coefficients (r) from
highest to lowest as follows: (A) -0.9, -0.5, 0, 0.6; (B) 0, -0.9, -0.5, 0.6; (C) 0, -0.5, 0.6, -0.9; (D) None of the above.

When asked this question, Student A selected answer A (-0.9, -0.5, 0.0, 0.6). After finding that her answer was
wrong, she said, The value of a correlation coefficient is similar to a numerical value, isnt it? Why is my answer
wrong? Student B also selected answer A, and her think-aloud utterances indicated that she thought all the other
answers were nonsense. When her SALS feedback indicated that her answer was wrong, she felt confused and said,
What? Is that wrong? Let me think for a while. Yeah! I think my answer is correct! Is it wrong? The student
then turned around and looked at the researchers. The answers from both students and their responses to the
corresponding items in the TDICM pre-test results showed that they had probably analogized the concept of
numerical value to the concept of correlation coefficient and had applied the misconception that a positive
correlation is stronger than a negative correlation. The learning processes of the two students demonstrates that this
188
learning phase can help students to become aware of their existing ideas, particularly their misconceptions, which is
the first and most important step in facilitating cognitive conflict.

The interview results also indicated that the context and presentation of the questions were important for cognitively
engaging the students to answer the questions and then to externalize them so that they became aware of their own
concepts or misconceptions. Further, since the students were cognitively engaged in the question situation, once they
discovered that their existing concept (e.g., numerical value) were wrong or unsuitable for explaining the new
concept (e.g., correlation coefficient), they eagerly sought more information. For example, Student A indicated that
the questions were related to situations she had encountered daily life. Therefore, when she found that her response
was wrong, she was puzzled. She also stated, In fact, I cannot believe that my answer is wrong of course, I want
to know why my response is wrong.

Reflection phase: The purpose of this phase was to guide students to answer questions about concepts raised in the
Externalization phase through simulation-based exploratory activities and also to demonstrate how the answers
differed from their own ideas, which were externalized in Externalization phase. For example, in the Reflection
phase of the fifth learning activity, the learning guide instructed the student to use the multiple windows function to
manipulate and compare scatter plots with different correlation coefficients (r =-0.9, -0.5, 0, and 0.6) and then to
explore the possible relationships among correlation coefficients(r) and corresponding scatter plots. During the
manipulation and comparison, Student A said, Let me see! In the figure (scatter plot) wherer is -0.9, all data points
fall almost on a line.In contrast, in the figure (scatter plot) where r is 0, the data points are so scattered. Is there a
rule? She then clicked the re-sampling buttons in the four windows and murmured to herself, It looks like
that...so I think my previous answer (in Externalization phase) may be wrong.

The learning process for Student B was more difficult. Although she also understood how the scatter plot with r =-
0.9 differed from the scatter plot with r =0, she did not realize that the positive and negative correlation coefficients
had different correlation directions. She said with a pointing finger, This (the scatter plot with r =-0.5) looks like
that (the scatter plot with r =0.6). Because she was unable to answer the question sets correctly, the SALS asked
her to compare the variations between the four scatter plots, which she again manipulated with different correlation
coefficients. After carefully comparing the four scatter plots, she said, Okay! Ive got it. According to the think-
aloud protocol, she thought that the direction of the scatter plot with r =-0.9 is like that with r =-0.6 but is in
contrast with the one with r =0.5. These observations of the student learning processes suggest that the guided
exploratory activities in this learning phase help students to reflect on their existing concepts.

Construction phase: In this phase of the simulation, the SALS learning guide clearly described the target concept and
the manipulation procedures so that the simulation could lead students to construct the correct correlation concepts.
For instance, after reading the definition of correlation strength, Student A referred to the SALS learning guide and
slowly moved the bar of the correlation coefficient from r =1 to r =-1 while observing the corresponding change in
the scatter plot and inferring the degree of correlation. After repeatedly moving the bar, she finally discovered the
trend of the change. She found that the scatter plots for the same correlation coefficient may have different patterns
but similar trends. She said in thinking aloud that when the value of r is changed to 1 or -1, the pattern of data points
in the plot would be more like an oblique line.

Next, Student A was asked by the learning guide to think about the characteristics of the scatter plots with r =1 or r
=-1 based on her earlier observation. She answered correctly. Her think-aloud protocol revealed that she thought that
the scatter plot is like a line when r =1 or r =-1, and the more that the scatter plot resembles an oblique line implies
a stronger correlation. Therefore, she thought that the correlation degree is the strongest when r =1 and r =-1.
Besides, during the interview, Student A said that the definition of the target concepts provided by learning guide is
useful for her to interpret her manipulation results and to understand the concepts. These student learning processes
and interview results demonstrate that this learning phase can help students to understand statistical concepts by
giving them sufficient knowledge of the target concepts as well as exploratory activities.

Application phase: In this phase, the students engaged in two problem situations to enhance their understanding of
concepts and to examine the degree of conceptual change. In the first problem situation, students were shown a
scatter plot with r =0.6 and asked to change it into a scatter plot with a stronger correlation degree and a negative
direction by removing the data points within it. Solving this problem required an understanding of the concept of
correlation degree and the conceptual relationship between a correlation coefficient and a scatter plot. Only relying
189
on a textual definition would have been insufficient. For example, although Student A indicated that she understood
the meaning of correlation degree, she was still confused when she first undertook this task. She murmured,
Uhwhat should I do? Uhthe more the figure (scatter plot) is like an oblique line, the stronger the correlation
degree it has. And when the correlation coefficient is negative, the figure (scatter plot) tends to be a line from the
higher left to the lower right. So, I should move the data points first, and thenAfter moving several data points
and rearranging the plot, she successfully completed the activity. These observations demonstrate that, although
Student A initially had no clear idea how to complete the task, she could gradually integrate and apply the statistical
concepts she had learned. Therefore, this task not only confirmed her understanding, it also helped her to construct a
more elaborate conceptual structure. Moreover, the scatter plot presented in this task differed from those in the
earlier learning activities, which showed that Student A could apply her constructed correlation concepts to solve an
unfamiliar problem. Therefore, her successful completion of this activity demonstrated her conceptual understanding
of correlation degree and direction. The second task was a situation-based multiple-choice test. Because the situation
was unfamiliar to the students, they were required to elaborate on their constructed statistical concepts. Both students
A and B correctly answered the question by applying the concepts they had learned.

Examining and elaborating students conceptions was the last step of SALS. Unlike the guided learning in previous
phases, this phase focused on enabling students to apply the correlation concepts in different situations and in
different ways but with minimal support, thus requiring them to construct more complete concepts. Depending on
their performance in this phase, SALS could also determine whether the students understood the target concept by
examining whether they could apply the constructed concepts to other novel problems. Students who were unable to
do so were instructed by the SALS to return to the Construction phase to relearn the concept. The interviews
indicated that this design increased the confidence of the students in learning and using statistics. For example, after
completing the Application phase activity, Student A said that the statistical concepts had many daily life
applications while Student B mentioned that statistics was not as daunting as she had expected.


The contributions of the simulation-based learning environment

Three methods (problem situations, appropriately structured guided learning, and contiguity principle) were used in
SALS to promote active learning, to reduce intrinsic cognitive load, and to reduce extraneous cognitive load when
learning with DLMRs during the Reflection and Construction phases. While earlier sections on students learning
processes during the two phases showed how these methods work and their effectiveness, this section describes
students perceptions about these methods.

First, the interview data revealed that presenting questions before the exploration activities enhanced DLMRs
learning. For example, Student B stated in the interview, The questions (presented before the manipulation)
stimulate me to pay attention because I want to know the answers. Additionally, guiding students to explore
DLMRs systematically reduced the intrinsic cognitive load of DLMRs learning. For example, Student A said, The
illustration of the second phase (Reflection) is very clear. Even though my knowledge of these symbols and figures is
not sufficient, I can learn what I should according to this (learning guide). Finally, the application of contiguity
principle in the design of DLMR functions was effective in reducing extraneous cognitive load. For example,
Student A said in the interview, Learning with simulations is very complex. I must pay attention to the figures
(scatter plots), which have many numbers (the values of r, x, y, and N) and relationships. Comparing different
manipulations is even more difficult. it (multiple windows function) is very useful since I dont need to remember
these figures and numbers.


Limitations of SALS

Although the evaluation results showed that the two students corrected their most correlation misconceptions and the
designs of the simulation-based learning environment could achieve their objectives, some limitations of SALS were
revealed by the think-aloud protocols, interview data and learning behaviors. For example, the students indicated that
some of the texts in the learning guide were difficult to understand and that the SALS response time was sometimes
slow. The two faults may reduce learning interest. For instance, student B complained when she manipulated
DLMRs, Why didnt the values of r change? I moved the bar (of r)quick, quick I hope that I can finish the
learning activity more quickly. Further, although the learning guide in the Reflection phase provided clear hints for
190
helping students learn statistical concepts by manipulating and observing DLMRs, the researcher found that, in the
third and seventh learning activities, Student B could not understand how representations from her manipulation
results are related and, thereafter, continued directly to the next phase. This may have affected her subsequent
learning, and she thus failed to correct two of her misconceptions. These problems should be considered in future
versions of SALS.


Conclusions and recommended future works

Although understanding and applying statistical concepts are essential abilities (Gal, 2002; Garfield & Ben-Zvi,
2007; Mills, 2002), students often have statistical misconceptions (Castro Sotos et al., 2007; Shaughnessy, 2007).
This study designed and developed Simulation Assisted Learning Statistics (SALS), a simulation-based CAL for
eliminating misconceptions about the statistical concept of correlation. Two important elements were considered in
the design and development of the SALS. One is the learning model based on cognitive conflict theory that includes
four learning phases, Externalization (making students aware of their implicit concepts), Reflection (guiding students
to find contradictory information and assisting them in reflecting on their existing concepts and misconceptions),
Construction (supporting students in constructing correct statistical concepts), and Application (examining the
change in misconceptions and helping students to transfer and apply the newly learned concepts). Another is the
design principles of a simulation-based learning environment based on cognitive load literature, including Inciting
germane cognitive load by expectancy-driven methods, Decreasing intrinsic cognitive load by presenting the
learning content in a structure way and Reducing students extraneous cognitive load with contiguity principle.

The results of the formative evaluation indicated that students substantially reduced their correlation misconceptions
after learning with SALS. Closely examining student learning processes and interview results also revealed that the
learning model of SALS and the design principles of simulation-based learning environment could achieve their own
objectives.

Based on the results of this study, the following areas for future research are recommended. First, given the
limitations of the SALS prototype, such as the difficult language used in the learning guide and the slow response
rate of SALS, further work is necessary to revise the language in the learning guide area of some activities and to
improve the SALS software. Further, to emphasize the manipulation results, a further question (e.g., What are the
features of your manipulation results?) should be included in future versions of SALS to remind students to
carefully observe and reflect on their manipulation results in the Reflection phase. Third, while simulation tools have
been widely applied in statistics education (Garfield & Ben-Zvi, 2007; Mills, 2002; Morris, 2001; Morris et al.,
2002), studies by Morris et al. (2002) and Mills (2002) suggest that further empirical studies are warranted to
evaluate the effectiveness of such tools for learning. This study only selected two undergraduate students who had
more misconceptions to be the study cases and conducted a formative evaluation to explore the merits and limitations
of SALS (prototype version). Based on the results of this study, Liu, Lin, & Kinshuk (in press) further conducted a
mixed method (embedded experiment model) with a larger number of samples to verify the effectives of SALS (the
revised version) and to investigate whether SALS can achieve the objectives that have been set for it. Further details
in that study can be found in Liu et al. (in press).

Fourth, although the formative evaluation demonstrated the potential of the three design principles of simulation-
based learning environment for learning statistics, further experiments should examine the effects of each of these
design principles. Finally, whereas the learning focus in this study was correlation, future works could test other
statistical concepts to confirm the positive effects of both the Learning Model and SALS.


Acknowledgements

The authors would like to thank the National Science Council of the Republic of China, Taiwan, for financially
supporting this research under Contract No. NSC 94-2520-S-008-002. The author would also like to thank the
students who participated in this study. Finally, the author would like to thank the editor of ET&S and the
anonymous reviewers of this paper for their kind assistance and helpful suggestions.


191
References

Ainsworth, S. (1999). The functions of multiple representations. Computers & Education, 33(2-3), 131-152.
Ainsworth, S. (2006). DeFT: A conceptual framework for considering learning with multiple representations. Learning and
Instruction, 16(3), 183-198.
Castro Sotos, A. E., Vanhoof, S., Van den Noortgate, W., & Onghena, P. (2007). Students misconceptions of statistical inference:
A review of the empirical evidence from research on statistics education. Educational Research Review, 2(2), 98-113.
Chan, C., Burtis, J., & Bereiter, C. (1997). Knowledge building as a mediator of conflict in conceptual change. Cognition and
Instruction, 15(1), 1-40.
Cohen, S., Smith, G., A., Chechile, R., Burns, G., & Tsai, F. (1996). Identifying impediments to learning probability and statistics
from an assessment of instructional software. Journal Educational and Behavioral Statistics, 21(1), 35-54.
Cumming, G., & Thomason, N. (1995). Designing software for cognitive change: StatPlay and understanding statistics. Paper
presented at the World Conference on Computers in Education VI. WCCE 95, J uly 23-28, Birmingham, United Kingdom.
Davison, G., Vogel, R., & Coffman, S. (1997). Think-aloud approaches to cognitive assessment and the articulated thoughts in
simulated situations paradigm. Journal of Consulting and Clinical Psychology, 65(6), 950-958.
Gal, I. (2002). Adults' statistical literacy: Meanings, components, responsibilities. International Statistical Review, 70(1), 1-25.
Garfield, J ., & Ben-Zvi, D. (2007). How students learn statistics revisited: A current review of research on teaching and learning
statistics. International Statistical Review, 75(3), 372-396.
Hewson, P. W., & Hewson, M. G. A. (1984). The role of conceptual conflict in conceptual change and the design of science
instruction. Instructional Science, 13(1), 1-13.
Kadijevich, D., Kokol-Voljc, V., & Lavicza, Z. (2008). Towards a suitable designed instruction on statistical reasoning:
Understanding sampling distribution with technology. Paper presented at the ICMI Study 18 Conference and IASE 2008 Round
Table Conference, J une 30 - J uly 4, Monterrey, Mxico.
Lee, G., & Kwon, J . (2001). What do we know about students' cognitive conflict in science classroom: A theoretical model of
cognitive conflict process. Paper presented at Annual Meeting of the Association for the Education of Teachers in Science,
J anuary 18-21, Costa Mesa, CA.
Limon, M. (2001). On the cognitive conflict as an instructional strategy for conceptual change: A critical appraisal. Learning and
Instruction, 11(4-5), 357-380.
Liu, T. C. (2007). Teaching in a wireless learning environment: A case study. Educational Technology & Society, 10(1), 107-123.
Liu, T. C., & Lin Y. C. (in press). Developing two-tier diagnostic instrument for exploring students statistical misconceptions:
Take Correlation as the example. Bulletin of Educational Psychology.
Liu, T. C., Lin, Y. C., & Tsia, C. C. (2009). Identifying misconceptions about statistical correlation and their possible causes
among high school students: An exploratory study using concept mapping with interviews. International Journal of Science &
Mathematics Education, 7(4), 791-820.
Liu, T. C., Lin, Y. C., & Kinshuk. (in press). The application of Simulation Assisted Learning Statistics (SALS) for correcting
misconceptions and improving understanding of correlation. Journal of Computer Assisted Learning.
Lowe, R. K. (1999). Extracting information from an animation during complex visual learning. European Journal of Psychology
of Education, 14(2), 225-244.
Mayer, R. E. (2002). Understanding conceptual change: A commentary. In M. Limon & L. Mason (Eds.), Reconsidering
conceptual change: Issues in theory and practice (pp. 101 111). Dordrecht: Kluwer Academic.
Mayer, R. E., & Moreno, R. (2003). Nine ways to reduce cognitive load in multimedia learning. Educational Psychologist, 38(1),
43-52.
Mayer, R. E. (2005). Principles for reducing extraneous processing in multimedia learning: Coherence, signaling, redundancy,
spatial contiguity, and temporal contiguity principles. In R. E. Mayer (Ed.), Cambridge Handbook of Multimedia Learning (pp.
183200). New York: Cambridge University Press.
Meletiou-Mavrotheris, M. (2004). Technological tools in the introductory statistics classroom: Effects on student understanding of
inferential statistics. International Journal of Computers for Mathematical Learning, 8(3), 265-297.
Mevarech, Z. R., & Kramarsky, B. (1997). From verbal descriptions to graphic representations: Stability and change in students
alternative conceptions. Educational Studies in Mathematics, 32(3), 229263.
192
Mills, J . D. (2002). Using computer simulation methods to teach statistics: A review of the literature. Journal of Statistics
Education, 10, Retrieved April 28, 2009, from http://www.amstat.org/publications/jse/v10n1/mills.html.
Morris, E. (2001). The design and evaluation of link: A computer-based learning system for correlation. British Journal of
Educational Technology, 32(1), 39-52.
Morris, E. J ., J oiner, R., & Scanlon, E. (2002). The contribution of computer-based activities to understanding statistics. Journal
of Computer Assisted Learning, 18(2), 114-124.
Shaughnessy, J . M. (2007). Research on statistics learning and reasoning. In F. K. Lester (Ed.), The Second Handbook of
Research on Mathematics (pp. 9571010). Reston, VA: National Council of Teachers of Mathematics (NCTM).
Strike, K., & Posner, G. (1985). A conceptual change view of learning and understanding. In L. West & R. Hamilton (Eds.),
Cognitive Structure and Conceptual Change (pp. 211-232). London: Academic Press.
Sweller, J ., van Merrienboer, J . J . G., & Paas, F. (1998). Cognitive architecture and instructional design. Educational Psychology
Review, 10(3), 251-296.
Tsui, C. Y., & Treagust, D. F. (2003). Genetics reasoning with multiple external representations. Research in Science Education,
33(1), 111-135.
van Somersen, M. W., Barnard, Y. F., & Sandberg, J . A. C. (1994). The think aloud method: A practical guide to modelling
cognitive processes, London: Academic Press.
Wouters, P., Paas, F., & van Merrienboer, J . J . G. (2008). How to optimize learning from animated models: A review of
guidelines based on cognitive load. Review of Educational Research, 78(3), 645-675.
Yin, R. (1994). Case study research: Design and methods (2
nd
Ed.), Beverly Hills, CA: Sage Publishing.
Yu, C. H., Behrens, J . T., & Anthony, S. (1995). Identification of misconceptions in learning central limit theorem and evaluation
of computer-based instruction as a remedial tool. Paper presented at the Annual Meeting of American Educational Researcher
Association, April 18-22, California, America.

También podría gustarte