Information Technology and Management (2020) 21:15–28

https://doi.org/10.1007/s10799-019-00299-9

Knowledge representation for computational thinking using

knowledge discovery computing
Youngseok Lee1 · Jungwon Cho2

Published online: 13 May 2019

© Springer Science+Business Media, LLC, part of Springer Nature 2019

Abstract
Modern society needs to think of new approaches for solving problems with computing. Computational thinking is the
process of abstracting and automating a variety of problems using computational technology. A system that expresses, man-
ages, and processes knowledge such as computational thinking is called a knowledge-based system. This paper proposes to
examine students’ knowledge about computational thinking when they want to develop a Python project, and the correlation/
association between these concepts. To achieve our goal, a field study was designed and data were collected from a com-
puter programming lecture. Through this data analysis, we try to identify the factors through the correlation between data
and clustering technique in order to express and discover the knowledge about the learner’s computational thinking. For the
verification of the factors identified, we analyzed the correlation between computational thinking and the pre- and post-test
results of the LightBot. In addition to the regression analysis of the proven factors, the probability of the research model was
analyzed through the structural equation to process the knowledge discovered. In this paper, we present various problems in
the domain of programming education and analyze the means to diagnose and improve knowledge based on computational
thinking by finding various problem-solving methods. To pre-examine the learner; he/she was diagnosed using a test paper
and the LightBot execution test. We checked the learner’s current knowledge state by analyzing the correlation between the
test site and the results of the LightBot. To analyze the level of knowledge improvement of learners, we designed an experi-
ment to analyze the correlation between learning and the actual test results through a system that applied the problem-solving
learning method. An analysis of the experimental results demonstrated that there was a correlation between the test results
for a learner and the pre-test results of the LightBot. Additionally, the group mean scores of the learners who learned as per
the proposed technique were observed to be significant. During this process, we analyzed the effects of problem-solving and
system application on academic achievement through factor analysis, regression analysis, and structural equation modeling.
The ability to pinpoint various problem scenarios and solve problems more effectively using computational technologies will
become more important in future. For this purpose, applying our proposed technique for deriving and improving knowledge
based on computational thinking to software education will induce the interest of students and increase the learning effect.

Keywords Knowledge representation · Knowledge discovery · Problem solving · Computational thinking · Knowledge-
based tutoring system

1 Introduction
* Jungwon Cho A knowledge-based system is an artificial intelligence sys-
[email protected]
tem that is used to solve complex problem, and it allows
Youngseok Lee knowledge to be represented [1]. Knowledge-based systems
[email protected]
are mostly used in problem-solving procedures, and they
1
KNU College of Liberal Arts and Sciences, Kangnam solve problems according to expert knowledge. In knowl-
University, 40 Gangnam‑ro, Giheung‑gu, Yongin‑si, edge-based systems, it is important to represent and manage
Gyeonggi‑do 16979, South Korea the knowledge of experts. You can learn how to represent
2
Department of Computer Education, Jeju National your own knowledge through software education [2].
University, 102 Jejudaehak‑ro, Jeju‑si, Jeju‑do 63243,
South Korea

13
Vol.:(0123456789)
16 Information Technology and Management (2020) 21:15–28

In the era of the current Fourth Industrial Revolution, and improve knowledge based on computational thinking
software is expected to become the center of value creation through the process of presenting various problem scenarios
and innovation. In accordance with these changes, software and finding various problem-solving methods for learning.
education is being improved in many countries [3]. Recently, This study aims at analyzing the effects of these types of
many universities are providing software education to all education on diagnosing and improving knowledge and ana-
students so that they can design and create software using lyzing the effects of this type of education on satisfaction
the knowledge of their major subject. This direction of of learners.
software education improves logical and creative thinking
through the process of ideation for problem-solving based
on computational thinking, resulting in people coming up 2 Related work
with their own solutions [4].
Computational thinking refers to the ability of solving 2.1 Knowledge discovery
problems through automated techniques by recognizing and
abstracting problems; this recognition and abstraction of Software education involves learning a programming lan-
problems are based on the basic concepts and principles of guage that allows one to communicate to the computer and
computational technology. Through software training, com- learn to order and check the results of the computer as one
puter specialists can help students shape their ideas using thinks. Currently, we can distinguish professional program-
computational technology in their major subject field and ming education to train software developers and universal
thus enable them to communicate with experts in various programming education to improve upon computational
fields. Computational thinking is computer-based think- thinking and problem-solving ability. In this study, we focus
ing; therefore, it can be developed and improved through on universal programming education to train talented people
software education based on educational programming lan- with computational thinking and problem-solving ability for
guages [5]. the preparation of the future society of the Fourth Industrial
Although there are various types of educational pro- Revolution era [9].
gramming languages for software education, it is desirable Programming finds ways to solve problems and resolve
to choose languages that are easy to learn and extend to them in the course of solving a given problem; it gradually
various types of application programs and additionally are completes a task by iterating and combining these processes.
of interest to non-computer-science students [6]. Python is From this perspective, various skills such as information
relatively easy to learn programming, and its graphics pro- processing ability, procedural thinking ability, problem-
cessing capability is simple; therefore, it is appropriate to be solving ability, logical thinking ability, and reasoning abil-
learned initially by beginners [7]. Python is also useful to ity are improved through programming [10]. To resolve this
develop various apps and web forms; therefore, it is highly problem, it is necessary to start with an easy problem that
likely to be used as a programming language for convergence one can solve and to go on further so that one can solve vari-
education. After analyzing the current situation both at home ous difficult and complicated situations [11].
and abroad, Python can be judged to be a suitable educa- To improve such computational thinking, it should be
tional programming language for computer specialists [8]. possible to utilize various knowledge on a computer basis.
However, in many universities, software education for Such knowledge can be roughly classified into three types
non-entrepreneurs is proceeding in a manner similar to [12]. In the first type, there is domain knowledge comprising
computer-oriented programming language-oriented educa- learning content objects, relationship information, explana-
tion and without considering specific educational goals and tion, examples, and problems between them. In the second
methods for the non-entrepreneurs [7]. Therefore, non-com- type, there is teaching knowledge as a necessary strategy
puter-science students face difficulties in learning to improve for the learning and teaching processes. Regarding the third
their computational thinking in the software curriculum. To type, it is the knowledge of learners accumulated from the
develop problem-solving skills, students are required to knowledge of the instructor. If such knowledge is structured
identify, present, and investigate real-world problems [8]. and stored in a repository, a system can be designed such
Additionally, the problems should be structured so that that knowledge can be selected and problems can be cor-
learning takes place in a series of processes that implement rectly recognized and solved. Additionally, the knowledge
problem-solving methods using an educational program- stored in the knowledge repository should be configured
ming language as a tool [9]. such that the results of various processes can be compared
In this paper, we present an example of problem-solv- [13].
ing-based software education that can enhance the think- In the system for utilizing such knowledge, an expert
ing ability of students while inducing interest in students to module, a teacher module, and learner module should be
facilitate software education. We analyze how to diagnose present. The expert module organizes the information that

13
Information Technology and Management (2020) 21:15–28 17

contains the knowledge that the learner wants to learn from customers. The customer base contains similar groups of
the instructor and manages the repository that stores it. The customers who share common tastes, interests or preferences
instructor module functions as a teacher; an example would (unsupervised segmentation), and we provide recommenda-
be different learning scenarios such as guided free play, tions of similar products. Thus, whenever you see statements
learning by doing, discovery learning, and mixed-initiative like, “People who like X also like Y” or “Customers with
dialog, and a real person who selects and orders teaching your browsing history have also looked at …,” similarity is
styles and teaching contents. The learner module comprises being applied [15].
the learner’s personal information, knowledge about the Clustering analysis techniques can obtain a meaningful
current learner’s area, and ability information about the data structure without dictionary information for knowledge
learner’s course. When such a knowledge-based system is representation and reasoning [16]. Clustering analysis tech-
constructed, the learner’s current competency is evaluated niques are applicable to almost all types of data and are
and the gap between the current level and the target compe- easy to apply because they do not need role definitions for
tency is analyzed. variables [16]. Clustering is the task of grouping together
After selecting the knowledge to be learned through this similar objects into several groups. For instance, n observa-
analysis and confirming the learning history and preference tions comprising p variables are divided into several groups
of the learner, the learning strategy is established. Fur- according to their similarity (proximity). A group of similar
ther, the appropriate learning material is selected, and the objects is called a cluster. Clustering aims to understand the
educational information to be used for the learner is then structure of the entire data by understanding the nature of
transferred to the expert and learner modules. At this time, each group, from which a set of rules is derived [16].
the expert module should have all the learning information To develop knowledge discovery and reasoning tech-
about the learning topic. The ability of the learner model to niques similar to human problem-solving procedures, we
provide all data on the learning topics needed to analyze and used non-hierarchical clustering methods [17]. K-means
evaluate the learner’s behavior is required [14]. clustering is a method of assigning each object to the clos-
In order that the learner module corrects the learning est center point [16]. First, we select K initial center points.
performance of the learner and to detect the error in the Second, each object is assigned to the cluster with the closest
learning activity according to the progress of learning, the center point, and the center point of the new cluster is then
expert module estimates the cause of error and suggests calculated. Third, the above steps are repeated until there
alternative knowledge about the problem. The instructor is no change in the allocation of each object. Finally, the K
module analyzes the proposed alternative knowledge and clusters are formed [16].
the learning strategy and preference of the learner. It further However, it is difficult to measure the non-similarity dis-
establishes a plan such as the order of learning and type of tance satisfying certain condition and determine the weight
learning contents suited to the correct learning method. This value that is generally used. In the case of K-average cluster
information is transferred to the expert learner modules and analysis, the results are poor if the setting of K is not appro-
further learning is conducted accordingly. When the current priate [16]. One disadvantage is that the interpretation of
learning process is completed, the next learning process is the results is ambiguous because there is no purpose given
selected through the learning performance diagnosis for cur- in advance.
rent learning. Further, the learning strategy is established,
and the procedure is presented to the learner. 2.3 Knowledge base for computational thinking

2.2 Knowledge representation and reasoning Computational thinking is the process of thinking that
includes defining and describing a problem so that a com-
Knowledge representation and reasoning are techniques for puter (person or machine) can perform it in an effective man-
analyzing or understanding complex problems and provide/ ner. An open-ended problem requires a comprehensive and
propose appropriate solutions. In knowledge discovery, it is meaningful solution based on several variables. To achieve
necessary to explain the knowledge in a form similar to the this, we need problem decomposition, knowledge represen-
process of human thought and to formulate the expressed tation, modeling, and algorithms found through computa-
knowledge in a manner such that inferences can be made in tional thinking [18, 19].
accordance with various conditions. Computational thinking involves decomposing the deci-
Knowledge representation is related to automated reason- sion process, taking into account the associated variables
ing. One of the primary purposes of explicitly expressing and all possible solutions, and choosing the correct decision
knowledge is the inference and assertion of new knowl- based on the corresponding method and problem limits [20].
edge [15]. Advertisers often want to serve online advertise- The general solution is to resolve the problem, identify the
ments to new consumers who are similar to their existing variables involved in representing the data, and generate the

13
18 Information Technology and Management (2020) 21:15–28

algorithm [21]. Common answers are generalizations and Students learn thinking strategies and domain knowl-
abstractions that can be used to solve the variants of the edge together in this learning. This form of problem-solving
problems that are derived from the original problem [20, learning originated in medical education and is now used in
21]. other fields as well [32]. The goal of problem-solved learn-
Computational thinking involves the following features ing is to develop flexible knowledge and to acquire effective
in the problem-solving process [22]: problem-solving skills, and to improve self-directed learn-
ing. Problem-solving is a form of active learning [33].
• Analyze and logically organize data Through group collaboration, students find out their cur-
• Data modeling, data abstraction, and simulation rent level of knowledge, what information they need to solve
• Organize user specific problem for computer help the problem, and how to approach it. The instructor should
• Identify, verify, and implement possible solutions develop students’ understanding, while at the same time,
• Automate solutions through algorithmic thinking instilling confidence in students to challenge the faculty and
• Generalize and apply this course to other problems. encouraging them for the same. Problem-solving learning
represents a paradigm shift that attempts to escape the tra-
Computational thinking is practical enough to be applied ditional teaching–learning philosophy centered on lectures.
to real-world problems such as “when keys are lost”, or The components that guide problem-solving learning are
“when determining the priority of pushed tasks”. It is also very different from traditional teaching methods [34].
applied in business, finance, engineering, and art [23, 24]. Problem-solving learning can improve the ability to solve
If a problem is solved first such as the problem-solving problems while understanding them, solving problem-solv-
method of the computer; or if the problem is identified and ing plans, practicing, revising, and re-practicing practical
structured, and an algorithm is introduced to solve it in a methods. In this study, we propose the following features
stepwise manner, it can be said that the problem is solved by of the problem that are required for the software education
computing accident [24]. For example, in the case when you based on the problem-solving learning based instructional
are looking for a lost key: “If the key is not in the room, look programming language education stage in various previous
inside the car. If it is not in the car, look in the coat pocket. studies [34].
If you cannot find it elsewhere, you can create a new key.”
The solution has a structure similar to the “If, else if, else” • Problems that describe or solve real-world situations
construct in computer programming languages [23, 24]. • Problems with more than one solution
Computational thinking is attracting attention as a way • Problems that can transfer knowledge
of new creative thinking. Its role is to help people solve • Motivating and challenging issues
problems across all fields easily and using a new approach. • Problems with expansion and disassembly
Computational thinking facilitates in the collection and
analysis of data and parallelizing problems to solve multiple Generally, the programming process is similar to the pro-
problems simultaneously [25]. In the modern society where cess of problem-solving. Therefore, this study analyzes the
computers dominate our lives entirely, The ability to identify research on problem-solving learning in various fields and
various problem situations and solve problems using com- applies it to software education.
puting technology will become increasingly important [26].
This is because computers are used in various fields such as
in design programs that use software to operate businesses, 3 Experimental design
talking with software subcontractors, and even expressing
art with computer technology. In this situation, computeri- 3.1 The process of knowledge discovery
zation of problems in various fields can be applied not only
to more efficient and accurate solutions but also to other The knowledge discovery process is interactive, iterative,
problems with similar types [27]. and involves the following critical steps [35]. To discover
Problem-solving learning and problem-based learning are knowledge, the domain of the knowledge to be found is
similar [28, 29]. It is a learning method in which learning is searched, and the data is transformed after preprocessing
performed in a series of processes that present problems to based on the data sample [35]. Based on this, data mining
the learners, search for solutions to the problems presented, techniques are applied to analyze patterns, leading to the
and develop solutions through individual or collaborative discovery of knowledge through verification [35]. Figure 1
learning [30]. Problem-solving learning or problem-based shows a common knowledge discovery process [35].
learning is a learner-centered learning environment and In this paper, the knowledge representation process for
model in which learners are taught to solve the presented the proposed knowledge discovery process is shown in
practical problems [31]. Fig. 2. Based on the general knowledge discovery process,

13
Information Technology and Management (2020) 21:15–28 19

Fig. 1  Knowledge discovery

process

Table 1  Initial cluster centers of K-means clustering analysis

Cluster
1 2 3 4

Initial cluster centers

Computational thinking 10 3 9 6

conducted to analyze the discovery and relevance of knowl-

edge on computational thinking. A K-means clustering was
conducted to analyze the discovery and relevance of knowl-
edge in computational thinking. The clustering analysis
results shows that computational thinking has four elements:
pattern, automation, abstraction, and algorithm, and a com-
putational thinking score ranging from 0 to 12 was assigned
to each of these four domains. The results are shown in
Tables 1 and 2.
Through this analysis, we try to identify the factors
through correlation between the data and clustering tech-
niques to express and discover the knowledge about the
learner’s computational thinking. For the verification of the
factors identified, we analyzed the correlation between com-
putational thinking and the pre- and post-test results of the
LightBot. In addition to the regression analysis of the proven
factors, the probability of the research model was analyzed
through the structural equation to process the knowledge
discovered.
Fig. 2  The proposed knowledge discovery process

3.2 Proposed knowledge‑based tutoring system

we propose a knowledge-based tutoring system and deter-
mine a knowledge discovery technique to utilize it. Figure 3 shows the structure of a system that extracts knowl-
We collected data from learner’s data and questionnaires edge suitable for the learner to provide knowledge-based
using the proposed system. The data collected result from tutoring, provides customized learning, and evaluates it
the measurement of the computational thinking abilities of based on the learning results. To achieve this, the knowl-
151 students who attended computer programming in the edge structure database for customized and problem-solving
first semester of 2018, which was held at the beginning of learning is included as a main feature, additionally including
the semester [34]. An analysis of the clustering experimen- the structure of the learning system.
tal results show that computational thinking appears as an The interface module provides the learning contents and
element of pattern, automation, abstraction, and algorithm. evaluation paper provided by the system in a form that can
The computational thinking score of these four domains be used by the learners. It interprets the response results of
is distributed from 0 to 12 points, K-means clustering is the learners and returns them to the system [36].

13
20 Information Technology and Management (2020) 21:15–28

Table 2  ANOVA of K-means Cluster Error F Sig.

clustering analysis
Mean square df Mean square df

ANOVA
Computational thinking 148.856 3 .310 147 479.557 .000

The ANOVA table indicates which variables contribute the most to your cluster solution
Variables with large F values provide the greatest separation between clusters

Fig. 3  Proposed knowledge-

based tutoring system

The learner module extracts learner characteristics and

level from the competency-based learner model and sug-
gests appropriate learning contents and evaluation items for
the individual. The diagnostic module of the learner mod-
ule processes the learner results received from the inter-
face module and generates the learner model based on the
instructor and learner knowledge. To re-analyze the learner
information, the evaluation is conducted, and the appropri-
ate items necessary for the learner’s level of education are
observed and presented [36].
The teacher module provides a choice of personalized
learning to enhance the learner’s vulnerable characteris-
tics. This is based on the learner’s characteristics and level
extracted from the learner module. Evaluation items are
extracted considering academic level and characteristics
[36].
The expert module manages the knowledge bases of con-
tents, questions, and teaching methods. The inference engine
determines the method by which the items can be extracted
based on the knowledge base of contents. It is possible to
determine the degree of difficulty of the items generated
by each rule through the knowledge base of the items [36].

3.3 Application example of problem solving

method of proposed system

Figure 4 shows the result of summarizing the learning

procedures to perform problem-solving learning. After Fig. 4  Procedure for problem-solving learning

13
Information Technology and Management (2020) 21:15–28 21

introducing the basics of programming, the problem and Table 3  Correlation between system scores and post-test of the Light-
its background are presented to the students [37]. Students Bot and actual academic performance
identify the problems they need to solve, what exactly they System score Academic Light Bot
need to solve in the problems and determine if they can be performance execution
solved. If the student does not understand the problem, he/ result
she can go back to understanding the basic element; if he/ System score
she can identify and judge the problem, the learning contents Pearson correlation 1
and additional items to solve are learned [38]. Sig. (2-tailed)
If a problem can be solved, its solution is summarized N 151
and presented. However, if it cannot be solved, then it is Academic performance
extracted and studied again. Each problem is arranged con- Pearson correlation .733** 1
sidering class objectives, problem contents, and class time. Sig. (2-tailed) .000
At the beginning of a class, easy and simple problems are N 151 151
presented; students are able to solve complex problems grad- LightBot execution result
ually. Thus, the problem-solving learning process is repeated Pearson correlation .444** .389** 1
[37]. The basic lecture plan is similar to the programming Sig. (2-tailed) .000 .000
education for non-computer specialists proposed in the cur- N 143 143 143
rent study; however, the contents of education are modified
so that problem-solving techniques can be applied [38]. **Correlation is significant at the 0.01 level (2-tailed)

3.4 Experimental result Table 4  Independent sample test conducted for the total learner
Elements N Avg. SD t p
During problem-solving learning, students developed solu-
tions for various general and coding problems through Light Bot execution result
Python programming. In this process, students used the pro- Pre 150 16.09 3.311 −5.511 .000
posed system and received points for each completed task. Post 143 18.03 2.667
We set this score as the system score. In addition, students
attempted to solve various problems on LightBot, which is
an easy coding-based Flash game that teaches the concepts Consequently, the results of the lecture and the results of
of programming. The LightBot execution result is a record the lecture through the LightBot are highly related to each
of students’ attempts to solve such problems. LightBot will other. The results of the independent sample test are shown
be described in more detail in Sect. 4. in Table 4.
To analyze the knowledge based on the computational The t test result should be chosen such that the same vari-
thinking ability for the problem-solving learning thus config- ance is not assumed, and a one-way test is conducted to con-
ured, the results using the system, the results of the post-test firm the learning performance. The results of the proposed
of the LightBot performed at the end of the semester, and system are as follows. The average score is 16.09 (standard
the results of the analysis of the correlation with the final deviation 3.311) and 18.23 (standard deviation 2.667). Thus,
academic performance are shown in Table 3. it can be seen that effective learning is achieved and there is
The correlation between the performance on the system a significant difference (significance probability p < 0.001)
and actual academic performance showed a very strong with 95% reliability. In other words, there was a difference
positive correlation with a value of .733 (p < 0.01). There- between the results of the pre-diagnosis and the post-diag-
fore, higher the system score, higher the actual academic nosis, and it was confirmed that the learning outcomes were
performance. The results of the light bot performance test better when using the computational thinking-based learning
conducted at the end of the semester showed that the cor- recommendation system for the suggested problem-solving
relation between the system performance was .444 (p < 0.01) learning.
and the correlation with the academic performance was .389
(p < 0.01) Therefore, the results of LightBot, which can be
evaluated in the process of execution, can be said to have 4 Case study
an academic correlation. Therefore, the use of the proposed
system for problem-solving learning based on computational 4.1 Discovering learner’s knowledge
thinking is highly related to the actual academic perfor-
mance, and it can be concluded that the students’ computa- To measure the computational thinking power of students
tional thinking ability is improved. after applying software education for the validity of the

13
22 Information Technology and Management (2020) 21:15–28

proposed software education model, we used the Korean can teach the elements of abstraction and automation while
intellectual educational information service and the compu- cultivating a sense of space. LightBot can increase mission
tational thinking ability test [34]. The results of this study difficulty with commands such as move, turn left, turn right,
are as follows. It is recommended that learners use this pre- jump, fire on, function 1, and function 2. The difficulty of
test evaluation to diagnose the levels of early learners. How- the mission is increased by raising thinking power and not
ever, since this test is a pen and pencil test that can be used by learning new concepts. This is an appropriate tool to
to measure computational thinking, it may be difficult to improve computational thinking skills, even for those with
measure the automation element of computational thinking computer skills.
through actual abstraction; therefore, we use a tool called At the beginning of the semester, we unraveled the com-
LightBot. putational thinking point to diagnose computational think-
LightBot is a programming puzzle game that uses a game ing ability and made it possible to solve the LightBot with
mechanism based on programming concepts [39]. It allows a brief description of 60 min. The computational thinking
the game learner to get a real understanding of basic control ability presented in Tables 1 and 2 consists of four detailed
flow concepts such as procedures, loops, and conditional elements (pattern, automation, abstraction, and algorithm),
statements [40]. The learner uses mouse-click commands and its score is defined by the combined score of the four
to move the robot on the tiles, thereby solving each level elements. We sought to understand the students’ knowledge
and understanding the concepts necessary for computational status by analyzing the correlation between the four ele-
thinking [39, 40]. LightBot can be used by anyone older ments, the students’ computational thinking (CT), and the
than 8 years of age and is used as a learner diagnostic tool LightBot pre-tests conducted at the beginning of the semes-
because it is suitable for diagnosing the initial learning abil- ter. Table 5 shows the result of analyzing the correlation
ity of computer specialists [40]. between the results of the computational thinking ability
test and LightBot.
• Light Bot Step 1: Basic Step As a result of the correlation analysis, the pattern, auto-
mation, abstraction, and algorithm were found to have a
Basic game rules are explained and executed in the fol- weak but significant correlation (p < 0.05). Particularly,
lowing manner. When the user clicks on each step, the Light- the results of the diagnosis of the computer and LightBot
Bot move one step forward to the square cell and move to the showed a positive correlation of .383 (p < 0.01) and .325
destination. Arrows are a way to succeed when a light bot is (p < 0.01), respectively. Therefore, higher the results of the
activated and reaches its final point by arranging a one-shot, computational thinking ability, higher the diagnostic results
left-turn, right-turn, jump, and an activity to turn on a light of the LightBot, and vice versa.
appropriately.
4.2 Design of research model
• Light Bot Step 2: Procedure Step
In this study, we analyze the process of diagnosis and
After executing the basic step, a procedure step with a improvement of knowledge of computational thinking for
slightly complex task is presented. Procedures can be under- problem-solving learning. The purpose of this paper was
stood in terms of functions. Separating a repetitive task as to investigate the effect of problem-solving process on
a procedure allows simplification of the code by calling the academic performance. For this purpose, the following
procedure whenever required. This enables to isolate repeat- four variables were selected to measure the diagnosis and
ing patterns throughout the process. improvement of the knowledge of computational thinking
For example, suppose there exists a pattern: ‘for- for problem-solving learning [37, 38].
ward > forward > forward > light on > right turn’ that repeats
twice. Here, it is possible to solve the problem by recording • Problem-solving learning
the repeated operation in the procedure area called PROC1 • Lecture composition
and calling the procedure name twice from the MAIN area. • Lecture participation
• Academic performance.
• Light Bot Step 3: Loops
Among these four variables, the data were extracted from
At this stage, the user can learn repetition. If P1 is placed the actual homework of students, midterm exam, and final
in the PROC1 procedure area to call itself back, one can exam. Questionnaires on problem-solving, lecture compo-
perform infinite iterations through the recursive function. sition, and lecture participation were also distributed to the
This mission is performed in 3D space through the LightBot students. The hypotheses to be revealed in this study are as
game; therefore, the user can naturally learn functions that follows.

13
Information Technology and Management (2020) 21:15–28 23

Table 5  Correlation between Pattern Automation Abstraction Algorithm CT LightBot

the results of the computational
thinking ability test and Pattern
LightBot
Pearson correlation 1
Sig. (2-tailed)
N 151
Automation
Pearson correlation .281** 1
Sig. (2-tailed) .000
N 151 151
Abstraction
Pearson correlation .273** .248** 1
Sig. (2-tailed) .001 .002
N 151 151 151
Algorithm
Pearson correlation .193* .182* .319** 1
Sig. (2-tailed) .018 .025 .000
N 151 151 151 151
Computational thinking
Pearson correlation .583** .698** .709** .639** 1
Sig. (2-tailed) .000 .000 .000 .000
N 151 151 151 151 151
LightBot
Pearson correlation .253** .325** .247** .179* .383** 1
Sig. (2-tailed) .002 .000 .002 .028 .000
N 150 150 150 150 150 150

**Correlation is significant at the 0.01 level (2-tailed)

*Correlation is significant at the 0.05 level (2-tailed)

Hypothesis 1 Problem-solving learning will affect stu-

dents’ academic performance.

Hypothesis 2 Problem-solving lecture composition using

learning contents recommendation system will affect stu-
dents’ academic performance.

Hypothesis 3 Interactions among students, faculty, and the

system will affect students’ academic performance.

The study model is based on the relationship between

variables and prior studies and it is shown by Fig. 5.

4.3 Sampling and data collection

We conducted a satisfaction survey to prove a study apply-

ing a computational thinking-based learning system for the
Fig. 5  Research model
proposed problem-solving learning. We conducted this sur-
vey with the use of a system in the computer programming
course having lectures from March to June 2018. The data Second, descriptive statistics and cross-correlation matri-
analysis process of this study is as follows. First, the reli- ces between the variables, which are the basic data of the
ability of the measurement tool was tested using Cronbach’s structural equation modeling evaluation, were calculated.
coefficient. In the structural equation modeling analysis, if the normal

13
24 Information Technology and Management (2020) 21:15–28

distribution assumption is severely violated, it may lead to and the appropriate questionnaire items were extracted. The
distorted results in the maximum likelihood method. The total reliability of the extracted questionnaire items was very
kurtosis and the degree of kurtosis were investigated to ver- high (.914). The results of reliability analysis are shown in
ify that each variable follows a normal distribution, and the Table 7.
versatile normal distribution was verified [39]. A total of 20 questionnaires were constructed. Reliability
Third, the fit and the model of the model were estimated. of all the questionnaires exceeded .9, indicating a very high
Fourth, the direct and indirect effects of variables were con- reliability. The results of the factor analysis for each item are
firmed. Fifth, the path coefficient and significance of the shown in Tables 8 and 9.
final model were tested and the magnitude of the total effect
was analyzed. 4.4 Measurement and result
Amos 18.0 was used for the analysis of this study and
statistical significance was verified at significance level of To analyze the structural equation model, there is a two-step
.05. Table 6 shows the definitions of the variables and the approach is used. In this approach, the measurement and
actual variables used in this study. structural models are analyzed as a single model, and the
The results of this study are as follows. First, the results validity is verified in the one-step approach and measure-
of the study on the attitude and interaction were analyzed ment model stage. In this study, we analyzed the structural

Table 6  The definitions of variables and related studies

Variable Operational definition Related study

Problem-solving learning The degree of functional friendliness of problem-solving learning Paik [28], GAO [41], Jaechoon [42]
Attitude toward proposed system The degree to which the participant has interest in lecture and proposed Youngseok and Jungwon [36], Ku
system [33], Jaechoon [42]
Interaction The degree of process and method that are related to each other in com- LIU [43], Junghwan [44],
munication between teacher and learner, user and system

Table 7  Reliability analysis results

Item Mean SD N Scale mean if Scale variance if Corrected item-total Squared multiple Cronbach’s
item deleted item deleted correlation correlation alpha if item
deleted

No. 1 3.93 .849 151 74.15 116.659 .721 .661 .906

No. 2 3.49 1.404 151 74.58 111.525 .577 .536 .911
No. 3 4.07 .892 151 74.01 116.393 .697 .604 .906
No. 4 4.20 .766 151 73.87 118.177 .711 .680 .907
No. 5 3.79 .803 151 74.28 120.895 .513 .465 .911
No. 6 4.12 .757 151 73.95 119.365 .645 .551 .908
No. 7 3.56 1.004 151 74.51 115.092 .673 .664 .907
No. 8 3.64 .989 151 74.43 113.927 .744 .769 .905
No. 9 3.70 .958 151 74.37 114.342 .749 .764 .905
No. 10 3.51 .979 151 74.56 114.514 .722 .741 .905
No. 11 3.84 .895 151 74.23 116.259 .701 .614 .906
No. 12 3.55 .862 151 74.52 116.064 .743 .649 .905
No. 13 4.40 .750 151 73.67 117.783 .753 .727 .906
No. 14 4.24 .846 151 73.83 118.006 .646 .605 .908
No. 15 3.97 .752 151 74.11 118.855 .682 .618 .907
No. 16 3.41 1.041 151 74.66 124.132 .232 .223 .918
No. 17 3.81 .700 151 74.26 137.929 − .476 .375 .927
No. 18 4.60 1.200 151 73.47 116.411 .492 .383 .912
No. 19 3.70 1.063 151 74.37 115.662 .604 .474 .909
No. 20 4.54 .746 151 73.54 126.077 .237 .238 .916

13
Information Technology and Management (2020) 21:15–28 25

Table 8  KMO and Bartlett’s test conformity by satisfying the condition of having a value of
.080 or less, CFI of .950 or greater, TLI of .950 or greater,
Kaiser–Meyer–Olkin measure of sampling adequacy .922
and RMSEA of less than .080.
Bartlett’s test of sphericity
The model consistency of confirmatory factor analysis
Approx. Chi square 1827.303
was estimated using the maximum likelihood estimation
df 190
method. The results showed that SRMR = .021, TLI = .955,
Sig. .000
CFI = .958, and RMSEA = .077 (LO: .064, HI: .090). Since
the number of cases is less than 250 and the number of
measurement variables is between than 12 and 30, the
relations of the research model using the two-step approach SRMR satisfies the condition of .021 or below .08. CFI
of Anderson and Gerbing [45], which analyzes the structural and TLI with values of .958 and .955, respectively, satisfy
equation model. the condition of being .92 or higher. Additionally, RMSEA
The conformity metrics of the measurement model were satisfies the criteria of model conformity by satisfying the
also based on the model conformance. The minimum and condition of being greater than .065 but below .080. The
maximum values of root mean square error of approxima- standardized and non-standardized coefficients of the set
tion (RMSEA), standardized root mean square residual measurement models are shown in Tables 10, 11, 12 and
(SRMR), Tucker–Lewis index (TLI) and comparative fit 13.
index (CFI) were analyzed [45]. In this study, the num- Table 11 shows the estimation results of the model.
ber of cases is less than 250, and the number of measure- The covariance and the correlation coefficient between
ment variables is between 12 and 30. Therefore, it can the potential variables seen in the set measurement model
be estimated that SRMR satisfies the criteria of model Fig. 3 are analyzed in Tables 12 and 13.

Table 9  Factor analysis result

Questions of questionnaire Factor
1 (problem- 2 (lecture 3 (lecture 4 (classroom
solving learn- composi- participa- environment)
ing) tion) tion)

No. 10 Have you improved your ability to self-program through problem-solving- .850
based learning?
No. 8 Did problem-based learning help me improve my computational or logical .835
thinking skills?
No. 9 Could problem-solving-based learning be unique to your programming .807
experience or method?
No. 7 Problem-solving Did you become interested in learning and motivated to .771
learn while you were learning?
No. 1 To what extent do you think you have gained new and useful knowledge .608
through this course?
No. 11 Did you actively participate in classes that use problem-based learning? .591
No. 19 How much do you think academic achievement has been achieved? .524
No. 12 What if I score a test? .517
No. 17 The progress of lectures, personal assignments, programming exercises, etc.. − .458
No. 13 What do you score on the teaching method? .780
No. 15 How do you rate your credit assessment method? .764
No. 14 What do you score on communication with professors? .757
No. 4 How do you rate your content? .723
No. 6 How about a variety of case-based problem-solving lectures? .584
No. 18 Recommended to take this course to colleagues or juniors .573
No. 5 What if I score a task? .496
No. 2 Compared to expectation of the semester….: .675
No. 20 How long have you been attending the lecture? .651
No. 3 What do you score the lessons you learned? .620
No. 16 What if you grade your classroom environment? .780

13
26 Information Technology and Management (2020) 21:15–28

Table 10  Confirmatory factor Model NPAR χ2 df SRMR TLI CFI RMSEA

analysis model compliance
index (n = 151) AVE LO90 HI90

CFA model 67 309.326 164 .021 .955 .958 .077 .064 .090

Table 11  Regression weights Estimate SE C.R. p

Academic performance ← Problem-solving learning .115 .068 2.689 .041

Academic performance ← Lecture composition .076 par_22
Academic performance ← Lecture participation .199 .109 1.822 .048

Table 12  Covariance result Estimate SE C.R. p

Problem-solving learning ↔ Lecture composition .492 par_15

Lecture participation ↔ Problem-solving learning .457 .075 6.126 ***
Lecture participation ↔ Lecture composition .484 par_17

***Significance at p < 0.01 respectively

Table 13  Correlations result We analyzed students’ previous knowledge state using the
Estimate
computational thinking ability test and the LightBot game.
The correlation between computational thinking and the sub
Problem-solving learning ↔ Lecture composition .723 elements pattern (0.583), automation (0.698), abstraction
Lecture participation ↔ Problem-solving learning .738 (0.709), and algorithm (0.639) were significantly high. The
Lecture participation ↔ Lecture composition .779 analysis also showed a high correlation between computa-
tional thinking and the LightBot game (0.383), which is a
useful tool to learn and understand computer programming
Covariance determines whether there is any correlation concepts and computational thinking.
between the two variables and degree of correlation. Cor- In addition, the analysis of the students’ group average
relation coefficient enables to know whether the discriminant scores according to the proposed technique showed that
validity is satisfied between the two factors. The correlation the scores significantly affect the LightBot test and the
coefficient is between .40 and .60 for the linguistic expres- actual academic performance. This study investigated how
sion according to the correlation coefficient, and the cor- to represent and discover students’ problem-solving abil-
relation is high when the correlation coefficient is between ity as knowledge, as well as how problem-solving learning,
.60 and .80. All the loads between the variables were above which enhances this knowledge, affects students’ academic
.457. It was found that the potential factors that were origi- performance. For each knowledge representation technique,
nally measured were appropriately measured by satisfying K-means clustering and various statistical analyzes were
convergent validity. conducted on the results of the application of the proposed
system.
4.5 Discussion In this study, how problem-solving learning affects
students’ academic performance was studied. The results
The purpose of this study was to analyze the process of showed that problem-solving learning directly affects stu-
diagnosis and improvement of knowledge based on compu- dents’ academic performance. The effect of the potential var-
tational thinking for problem-solving learning; and to ana- iable was .135, which showed a statistically significant effect
lyze the structural relation between problem-solving, lecture at significance level .001. The results of this study show
composition, and lecture participation. For this purpose, the that the ability of students to solve problems by improving
reliability and validity of the research model and measure- their own computational skills and logical thinking abil-
ment tools were verified and the structural relationship of the ity by applying various problem-solving methods has been
research model was analyzed. Statistical tests revealed that improved to improve their grades.
the causality between these potential variables was signifi- Additionally, it can be said that the structure of the lec-
cant and the validity of the study model was verified. ture applying the learning content recommendation system

13
Information Technology and Management (2020) 21:15–28 27

based on the problem-solving learning directly affects the 4. KERIS (2016) Research report KR 2016-4. http://lib.keris​.or.kr/
academic achievement of the students. Likewise, the inter- searc​h/detai​l/CATLA​B0000​00012​086. Accessed 25 Sept 2017
5. Ater-Kranov A et al (2010) Developing a community definition
action between students, faculty, and the system can affect and teaching modules for computational thinking: accomplish-
students’ academic performance. However, it is difficult to ments and challenges. In: Proceedings of the 2010 ACM confer-
say if the physical environment of the classroom affects the ence on Information technology education. ACM, pp 143–148
actual academic performance. However, it is necessary to 6. Jin SH, Shin S (2013) Case study and needs analysis on con-
vergence education in engineering colleges. J Eng Educ Res
consider the facility environment of the classroom for the 16:29–37
psychological incentives and effects of the students. 7. Python Software Foundation (2017) Python about. https​://www.
pytho​n.org/about​/. Accessed 5 July 2018
8. Guo P (2014) Python is now the most popular introductory teach-
ing language at top U.S. universities, BLOG@CACM. http://
5 Limitations and conclusions cacm.acm.org/blogs/​ blog-cacm/176450​ -python​ -is-now-the-most-
popul​ar-intro​ducto​r y-teach​ing-langu​age-at-top-us-unive​rsiti​es/
Computational thinking is emphasized as the most important fullt​ext. Accessed 12 June 2018
ability for everyone, regardless of their major subject, to pre- 9. Lee HW, Min HR, Yi KW (2008) A study on the improvable pro-
posal of general education curriculum of engineering college—a
pare for the current era of the Fourth Industrial Revolution. case of Seoul National University. J Eng Educ Res 11(3):24–32
In the future, it is predicted that people with talent who have 10. Kim SH (2015) Analysis of non-computer majors’ difficulties in
a superior ability of computational thinking ability and prob- computational thinking education. J Korean Assoc Comput Educ
lem-solving ability based on humanistic literacy are needed. 18(3):15–23
11. Duncan C, Bell T (2015) A pilot computer science and program-
Software education as a universal education for this pur- ming course for primary school students. In: Proceedings of the
pose should be structured so that it can be interesting to workshop in primary and secondary computing education. ACM,
solve the problem using computational technology, but many pp 39–48
universities still do not exceed the limits of programming 12. Sung Y (2017) Development of SW education model based on
HVC learning strategy for improving computational thinking. J
language education. Korean Assoc Inf Educ 21(5):583–593
In this study, we have proposed software education as a 13. Park SH (2016) Study of SW education in university to enhance
problem-centered form based on the Python language, which computational thinking. J Digit Converg 14(4):1–10
is widely used as an educational programming language in 14. Jeong I (2017) Study on the preliminary teachers’ perception for
the development of curriculum of the robot-based software edu-
college. We solved the problem by using a solution accord- cation in the universities of education. J Korean Assoc Inf Educ
ing to the presented problem scenario. Consequently, it was 21(3):277–284
found that students’ knowledge acquisition and academic 15. Jeon Y, Kim T (2015) The design and application of an expe-
achievement improved. Additionally, the correlation between rience-driven online software class based on creative problem
solving for cultivating the creative personality of the elementary
the problem-solving process and academic performance of informatics-gifted students. J Korea Elem Educ 26(4):477–494
the actual students was analyzed to obtain relevant results. 16. WIKIPEDIA (2018) Knowledge representation and reasoning.
In future, we plan to research on ways to improve com- https​://en.wikip​edia.org/wiki/Knowl​edge_repre​senta​tion_and_
putational thinking power and problem-solving ability reaso​ning. Accessed 28 July 2018
17. Tan PN, Steinbach M, Kumar V (2013) Data mining cluster analy-
based on this study, and on the effectiveness of software sis: basic concepts and algorithms. Introduction to data mining,
education by applying the software education subject to pp 487–533
real-world problems in non-major areas. In the future, I will 18. Liu D-R, Ke C-K (2007) Knowledge support for problem-solving
study research on ways to improve computational thinking in a production process: a hybrid of knowledge discovery and
case-based reasoning. Expert Syst Appl 33(1):147–161
power and problem solving ability based on this, and study 19. Jung HY (2014) An empirical study on information liberal educa-
on effectiveness of software education by applying software tion in university based on IT fluency and computational thinking
education subject to real life problems by non-major areas. concept. J Korea Soc Comput Inf 19(2):263–274
20. Chen G, Shen J, Barth-Cohen L, Jiang S, Huang X, Eltouhky M
(2017) Assessing elementary students’ computational thinking
in everyday reasoning and robotics programming. Comput Educ
References 109:162–175
21. Kim K, Kim H (2014) A case study on necessity of computer
1. Rajagopalan C, Baldev R, Kalyanasundaram P (1996) The role programming for interdisciplinary education. J Digit Converg
of artificial intelligence in non-destructive testing and evaluation. 12(11):339–348
Insight 38(2):118–123 22. Kim CW (2010) The application of computer education in the
2. Techopedia (2018) Knowledge-based system (KBS). https:​ //www. revise curriculum. J Educ Res Inst 12(1):41–55
techo​pedia​.com/defin​ition​/7969/knowl​edge-based​-syste​m-kbs. 23. Grover S, Cooper S, Pea R (2014) Assessing computational learn-
Accessed 30 Aug 2018 ing in K-12. In: Proceedings of the 2014 conference on innovation
3. García-Peñalvo FJ, Cruz-Benito J (2016) Computational thinking & technology in computer science education. ACM, pp 57–62
in pre-university education. In: Proceedings of the fourth inter- 24. Jun S (2017) Design and effect of development-oriented model for
national conference on technological ecosystems for enhancing developing computing thinking in sw education. J Korean Assoc
multiculturality. ACM, pp 13–17 Inf Educ 21(6):619–627

13
28 Information Technology and Management (2020) 21:15–28

25. Atmatzidou S, Demetriadis S (2016) Advancing students’ com- 36. Lee Y, Cho J (2015) Personalized item generation method for
putational thinking skills through educational robotics: a study on adaptive testing systems. Multimed Tools Appl 74(19):8571–8591
age and gender relevant differences. Robot Auton Syst 75:661–670 37. Lee Y (2018) Python-based software education model for non-
26. Basawapatna A et al (2011) Recognizing computational thinking computer majors. J Korea Converg Soc 9(3):73–78
patterns. In: Proceedings of the 42nd ACM technical symposium 38. Lee Y (2018) Analyzing the effect of software education applying
on computer science education. ACM, pp 245–250 problem-solving learning. J Digit Converg 16(3):95–100
27. Seo J (2017) A case study on programming learning of non- 39. Yaroslavski D (2018) LightBot. http://Light​Bot.com/. Accessed
SW majors for SW convergence education. J Digit Converg 20 June 2018
15(7):123–132 40. Gouws L, Bradshaw K, Wentworth P (2013) First year student
28. Paik S (2017) The effects of educational programming language performance in a test for computational thinking. In: Proceed-
with PBL (problem based learning) on logical thinking ability and ings of the South African Institute for Computer Scientists and
problem solving ability in elementary school environments, Mas- Information Technologists conference. ACM, pp 271–277
ter thesis. Korea National University of Education, Chung-Buk 41. Gao S et al (2016) Factors affecting the performance of knowledge
29. Myung HJ (2014) Effects of software special classon program- collaboration in virtual team based on capital appreciation. Inf
ming and creative problem solving capability, Master thesis. Han- Technol Manag 17(2):119–131
yang University, Seoul 42. Jo J et al (2016) A study on factor analysis to support knowledge
30. Kim B, Jeon Y, Kim J, Kim T (2016) Development and applica- based decisions for a smart class. Inf Technol Manag 17(1):43–56
tion of real life problem solving lesson contents based on com- 43. Liu H et al (2014) A model for consumer knowledge contribu-
putational thinking for informatics integrated-gifted elementary tion behavior: the roles of host firm management practices,
school students’ creativity. Korean J Teach Educ 32(1):159–186 technology effectiveness, and social capital. Inf Technol Manag
31. Ku J, Jeon Y, Kim T (2016) The development and application of 15(4):255–270
lesson contents based on the CT-CPS framework for improving 44. Junghwan L, Munkee C, Hwansoo L (2015) Factors affecting
the creative problem solving ability of elementary informatics smart learning adoption in workplaces: comparing large enter-
gifted students. J Korea Elem Educ 27(2):339–357 prises and SMEs. Inf Technol Manag 16(4):291–302
32. Ki JY (2018) A study on UX design process lecture based on 45. Anderson JS, Prussia GE (1997) The self-leadership question-
modified PBL (problem-based learning). J Korea Converg Soc naire: preliminary assessment of construct validity. J Leadersh
9(1):117–131 Stud 4(2):119–143
33. Ku JH (2017) Designing an app inventor curriculum for compu-
tational thinking based non-majors software education. J Converg Publisher’s Note Springer Nature remains neutral with regard to
Inf Technol 7(1):61–66 jurisdictional claims in published maps and institutional affiliations.
34. Lee Y, Cho J (2017) The influence of Python programming educa-
tion for raising computational thinking. Int J u- and e- Serv Sci
Technol 10(8):59–72
35. Dhiman AK (2011) Knowledge discovery in databases and librar-
ies. DESIDOC J Libr Inf Technol 31(6):446–451

13
Information Technology & Management is a copyright of Springer, 2020. All Rights
Reserved.