You are on page 1of 11

Asia-Pacific Edu Res (2020) 29(1):9–19

https://doi.org/10.1007/s40299-019-00442-8

REGULAR ARTICLE

A Content Analysis of Computational Thinking Research:


An International Publication Trends and Research Typology
Kai-Yu Tang1 • Te-Lien Chou2,3 • Chin-Chung Tsai4,5

Received: 16 September 2018 / Accepted: 9 March 2019 / Published online: 21 March 2019
 De La Salle University 2019

Abstract This paper aims to provide a comprehensive with traditional command-based tools, Scratch, Lego, and
analysis of publication patterns on computational thinking Python were identified as the emerging visual-based pro-
(CT) over two recent periods (period I: 2006–2012; period gramming languages. (3) Finally, the keywords related to
II: 2013–2018). Based on keyword analysis, a total of 3798 learning outcomes were classified based on the Bloom’s
(period I) and 7175 (period II) keywords were found. framework of three learning domains. First, knowledge and
Derived from the content analysis, a research typology of mental understanding are the main goals in the cognitive
two-period keywords was consolidated and framed domain; motivation and attitude are the main tasks in the
according to its attributes, including background settings, affective domain; and social and communication skills are
domain-specific factors, and learning outcomes. Main the central outcomes in the training of psychomotor ability.
findings show as follows: (1) Regarding the research Further discussions and research directions are provided.
background, students from secondary and higher education
are the main participants; and computer science, mathe- Keywords Computational thinking  Keyword analysis 
matics, and engineering are the major subjects. (2) As the Content analysis  International publication trends 
domain-specific factors, game and peer collaboration were Research typology
found to be the main pedagogies, while web-based and
face-to-face learning environments were almost equally
referred to in CT research settings. However, compared Introduction

Since Wing (2006) proposed the new concept of ‘‘com-


& Kai-Yu Tang putational thinking’’ (CT) and advocated that everyone
ky.nctu@gmail.com
develop the ‘‘ways to think like a computer scientist’’ from
1
Department of International Business, Ming Chuan kindergarten to higher education, the concept has now
University, 250 Zhong Shan N. Rd., Sec. 5, Shilin Dist., become one of the fundamental skills of the twenty-first
Taipei 111, Taiwan century (Kim et al. 2012; Shailaja and Sridaran 2015;
2
Graduate Institute of Digital Learning and Education, Shute et al. 2017). CT represents an attitude and skill for
National Taiwan University of Science and Technology, 43, solving problems, designing complex systems, and under-
Keelung Rd., Sec.4, Da’an Dist., Taipei 10607, Taiwan
standing human thoughts and behaviors that draw on
3
Department of Library and Information Science, School of concepts fundamental to computer science (Lye and Koh
Continuing Education, Fu-Jen Catholic University, 510,
Zhongzheng Rd., Xinzhuang Dist., New Taipei City 24205,
2014). Through programming practices, students learn to
Taiwan generalize abstractions, process information and detect
4 errors systematically, compose and decompose problems
Program of Learning Sciences, National Taiwan Normal
University, Taipei, Taiwan structurally, and think in iterative, recursive, and parallel
5 ways (Grover and Pea 2013). Moreover, students could
Institute for Research Excellent in Learning Sciences,
National Taiwan Normal University, 162, Section 1, Heping develop logical thinking skills and transfer them into
E. Rd., Da’an Dist., Taipei 106, Taiwan

123
10 K.-Y. Tang et al.

different contexts to face more challenging problems in effectively and efficiently with solutions that are reusable
real life (Voogt et al. 2015). in different contexts.’’ Finally, they categorized CT skills
According to the search from the Web of Science into six facets, including decomposition, abstraction,
(WoS), research reviews on CT have gradually increased algorithm design, debugging, iteration, and generalization.
since 2013. While some reviews have focused on K-12 to To provide a complementary viewpoint to previous
higher education, others have summarized the definitions, review papers, the present study selected an academic-fo-
assessments, and research agenda. Based on a narrative cused database from the WoS as the source of research
review, Grover et al. (2013) investigated the K-12 educa- data. A starting year of 2006 was set to line up with the
tion on CT in terms of environments and tools. They year of Wing’s (2006) original paper, and the review ended
concluded that educators needed to adopt graphical or in the middle of 2018 to provide the cutting-edge research
game-like programming tools to inspire students’ creativity status of CT. Last, the proposed typology was grounded on
and problem-solving skills for K-12 learners. Regarding a huge number of keywords to provide a comprehensive
the teaching strategies for CT education, Lye and Koh survey of the literature development. This bottom-up
(2014) surveyed 27 studies and found that CT was mainly approach aims to provide solid evidence of the research
taught through programming. They concluded that the trends.
scaffolding process is a key element in CT cultivation. The main foci of this current study are twofold. First, the
Using content analysis, two recent reviews provided more focus was to profile international publication patterns of
in-depth analyses of CT characteristics, practices, tools, CT research over two recent periods (period I: 2006–2012;
and models. Buitrago Flórez, et al. (2017) analyzed 92 period II: 2013–2018). According to the line chart of CT
papers from the ACM digital library and selected journal publications from 2006 to 2018, a slow growth trend in
publications. The authors highlighted the importance of terms of articles and authors occurs in the period I
peer-based collaborative environments and concept maps (2006–2012), followed by more obvious increasing trend in
for enhancing meaningful programming experiences. They period II (2013–2018). As shown in Fig. 1, the cutoff point
also concluded that programming can be used as a tool for was thus set between 2012 and 2013 to make a balance for
mind development, and pointed out some challenges faced the analyzed years of publications. A balanced number of
by early programmers, tools, and pedagogies used in years for the subsequent analysis and comparison led to a
schools. Similarly, Shute et al. (2017) searched for research lesser distortion of analysis. Moreover, this setting can
papers from various databases, such as ERIC, JSTOR, and benefit researchers to observe the potential changes in CT
Google Scholar. They selected 45 studies for review and publication patterns in these two periods, such as interna-
created a working definition. They defined CT as ‘‘the tional productivity. The second focus was to identify the
conceptual foundation required to solve problems most distinctive keywords as a typology for characterizing

Fig. 1 A line chart of CT


publications (2006–2018): the
2013-2018
amounts of journals articles and average papers: 80.9
authors. *The timespan of data average authors: 186.7
retrieval started from 2006 and
300 299
ends on August 10, 2018

250 2006-2012
average papers: 36.1 209
200 average authors: 79.7 190
164
144
150 124
114 129
94 102 101
100 96
76 75 73
47 57
48 45 47 45 48
50
22 34
24
10
0
2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018*

Arcle growth stascs Author growth stascs

123
A Content Analysis of Computational Thinking Research: An International Publication Trends and… 11

Table 1 The strategy of keywords used in the search


Step Keywords Results

1 TS = (‘‘comput* thinking’’) 136


2 TS = (‘‘comput* education’’ or ‘‘computer* science education’’) 394
3 TS = (‘‘comput* literac*’’) 321
4 TS = (abstraction or automation or decomposition or ‘‘thinking recursively’’) 318
5 TS = (‘‘programming’’ or ‘‘computer* program*’’ or ‘‘cod* thinking’’ or ‘‘algorithm* thinking’’ or ‘‘comput* intelligence’’) 838
6 Two original articles with 394 citing articles 396
7 #6 or #5 or #4 or #3 or #2 or #1 1611
8 Conducting a relevancy test for the search #7 715
In line with the education focus, the search of computational thinking was limited to the category of ‘‘education educational research’’ defined by
the Web of Science. Two original works published in computer science journals by Wing (2006, 2008) were included and treated as seeding
articles to include all their citing journal articles. Based on the setting of the search, research articles related to computational thinking from the
computer science domain to education and educational research were covered as much as possible. The timespan of the search period
(2006–2018) was set and conducted on August 10, 2018. Indexes = SSCI, SCI-expanded

CT research. This is a strategy for research review. ‘‘thinking recursively,’’ ‘‘programming,’’ ‘‘computer pro-
Adopted from earlier studies, researchers used to classify gram,’’ ‘‘coding thinking,’’ ‘‘algorithm thinking,’’ and
highly used keywords for presenting research themes and ‘‘computing intelligence.’’ In addition, a strategy for the
framework (Maida and Cheon 2014; Hsiao et al. 2015; use of keyword searching was applied as shown in Table 1.
Park et al. 2016). However, rather than common used The first five steps involved searching for papers that
wordings, this current study focus on selecting keywords included the keywords listed above. These multi-keyword
that can represent attributes of CT research. As such, two queries were to search for literature with the core concept
methods were adopted to achieve the research goals. While (i.e., computational thinking), its synonymous concepts/
the keyword analysis is an initial attempt to systematically terms (e.g., coding thinking), its practical competencies
profile all the keywords used in CT research, the following (e.g., abstraction), and all possible settings or approaches
content analysis is to identify the most critical keywords (e.g., computer education) step by step. The search was
for the comparison of CT development across the two limited to the category of ‘‘education educational research’’
periods. In sum, after a brief introduction, the following as defined by the WoS to be in line with the education
section illustrates the process of data retrieval and the focus. The truncation searching technique (*) was used to
details of the used methods. Finally, the main results are take in keyword variations. For example, ‘‘comput*
interpreted and discussed. thinking’’ returns results with ‘‘computational thinking’’
and ‘‘computing thinking’’; ‘‘comput* education’’ returns
results with ‘‘computer education,’’ ‘‘computing educa-
Data and Method tion,’’ and ‘‘computational education.’’ As shown in
Table 1, these five steps found 136, 394, 321, 318, and 838
The process of data inclusion is critical for conducting a research papers, respectively. Next, in step 6, two original
content analysis of literature reviews. The Web of Science articles of computational thinking (Wing 2006, 2008) were
(WoS) database, the major source of high quality indexed included along with all citing articles to capture a com-
academic journals that contains both the Sciences Citation prehensive and the most up-to-date developments in the
Index (SCI) and the Social Sciences Citation Index (SSCI) literature. As such, we set the data-retrieval time to begin
databases, was explored in this study. This paper used a from 2006 and to end on August 10, 2018. Based on the
systematic keyword search and a citation-based relevancy keywords search, a total of 1611 articles were retrieved (in
test to secure our research data. step 7).
First, the keyword definitions were based on four highly Second, the citation-based relevancy test was performed
cited papers, including two original works on computa- by a citation network analysis to control those documents
tional thinking authored by Jeannette Wing (2006, 2008), that have no citations and are not citing documents in the
and two recent research review papers (Grover and Pea dataset. This practice is conducted because the topic search
2013; Lye and Koh 2014). The keywords used for the in the WoS will search the article title, author-defined
query included ‘‘computational thinking,’’ ‘‘computer keywords, abstract, and Keywords Plus. However, Key-
education,’’ ‘‘computer science education,’’ ‘‘computer words Plus is a column with system-defined keywords,
literacy,’’ ‘‘abstraction,’’ ‘‘automation,’’ ‘‘decomposition,’’ and it may have less relationship with the interests of

123
12 K.-Y. Tang et al.

research. Based on the logic of citation use suggested by keywords between two periods was conducted for identi-
past reviews (Liu et al. 2017), those irrelevant documents fying the research trend among CT publications.
were excluded. As a result, a total of 715 articles remained
(step 8), representing a research pool of computational Content Analysis
thinking research published in high quality journals in the
recent decade. Meanwhile, all the bibliometric information The second method we used was content analysis. The
of the 715 articles was collected for the following keyword main idea of a content analysis is to categorize the scien-
and content analyses. tific findings to uncover the underlying dynamics concep-
tually and categorically, which is similar to grounded
Keyword Analysis theory (Glaser and Strauss 1967). Past research has sug-
gested that the use of content analysis helps researchers
The term ‘‘keyword’’ used in this paper is a general usage explore some prominent trends and pivotal insights that
to indicate the key information appearing in the titles, have not yet been found (Lin et al. forthcoming). As such,
keywords, and abstracts. Some researchers have suggested the content analysis adopted in this study could even pro-
that the wordings highlighted in the articles’ titles, key- vide scaffolds to build up a conceptual framework for the
words, and abstracts was considered as the most essential research community (Dyrbye et al. 2007).
but critical element to represent an article (Cook et al. All the two-period keywords were then content-ana-
2007). This idea is also identical with the topic search used lyzed within the context of computer science education.
in the WoS. Recently, keyword analysis has been applied First, to consolidate the most distinctive keywords, the
to explore over four decades of articles published in a process of induction was continued for several runs by two
renowned education journal (Zawacki-Richter and experienced researchers. A consensus was reached if the
Latchem 2018). keyword was selected by the two researchers, while those
As such, a keyword analysis was performed by keywords that failed to gain consent were resolved upon
decomposing the most valuable content of the corpus from discussion. Overall, the intercoder reliability for each
the 715 CT articles. First, all the wordings in the columns analysis was at least 85%, indicating that the induction was
of title, keyword, and abstract were gathered. Some stop adequately reliable.
words which are insignificant were excluded, such as Furthermore, each keyword was assessed based on its
‘‘and,’’ ‘‘at,’’ ‘‘in,’’ ‘‘on,’’ ‘‘above,’’ ‘‘under,’’ and so on. nature and was classified according to its various attributes.
Other common terms were also removed, for example For example, elementary, secondary, and teacher are the
‘‘computer,’’ ‘‘education,’’ ‘‘learning,’’ and ‘‘computational types of participants; arts, biology, chemistry, computer
thinking.’’ The process of scraping common terms is to science, and robotics are the different subjects studied in
reduce noise because those wordings are less discrimina- the research. The keywords with C/C??, R, Lego, Python
tive for identifying the main characteristics of CT research. were categorized as programming languages. While some
Second, the terms with similar or identical meanings in the keywords served as the background settings of a study (i.e.,
context of CT research were regarded as being the same. participants, subjects), others are domain-specific factors
For instance, ‘‘algorithm’’ and ‘‘algorithmic,’’ ‘‘APOS’’ (i.e., pedagogy, environment, programming lan-
and ‘‘APOS theory,’’ ‘‘computer literacy’’ and ‘‘computer guages/tools). In addition, following past research, key-
literate’’ are all treated as the same wordings. The APOS is words related to learning outcomes were divided into three
a procedure of action, process, object, and schema used in domains of learning (Bloom et al. 1956), namely cognitive,
computer education (Dubinsky and McDonald 2001). affective, and psychomotor outcomes. Finally, a typology
After the list of keywords was aligned, each keyword of 67 selected keywords in CT research was obtained as
was then matched and counted if it appeared in a certain shown in Table 2.
article. Note that each article may have multiple keywords;
however, those duplicated wordings used in a single article
were counted only once. Only the keywords appeared in Results and Discussions
the different article were accumulated. The use of unique
counting is helpful to compare how many times a sub- This study analyzed the CT studies from 2006 to 2018. The
ject/tool/concept or noteworthy practice in CT research, results of the most productive countries, influential articles,
which is represented as the keyword, has been addressed and active authors, as well as some distinctive keywords
among the 715 articles. Accordingly, two sets of keywords that representing main attributes of CT research typology
from the two periods of analysis were identified (3798 are presented and discussed below.
keywords from period I: 2006–2012, and 7175 keywords
from period II: 2013–2018). A further comparison of

123
A Content Analysis of Computational Thinking Research: An International Publication Trends and… 13

Table 2 A typology of selected keywords


Category Keywords Description

1. Participants Elementary, Higher education, Secondary, Teacher Background information that is often used in the CT
2. Subjects Arts, Biology, Chemistry, Computers science, Ecology, Engineering, research
Languages, Mathematics, Music, Physics, Robotics course
1. Pedagogy APOS, Constructivism, Game, Inquiry, Microworld, Pair Domain-specific factors—some unique features or
programming, Peer/Collaboration, Problem-based learning, Self- characteristics in the research context and setting
regulated learning, Simulation, Situated of CT
2. Environment Blended/Hybrid, Face-to-face, Web-based/Online
3. Programming Alice, Augmented reality, C/C??, Code.org, Java, JavaScript, Kodu,
languages/tools Lego, Python, R, Robots, Scratch
1. Cognitive Abstraction, Academic, Critical thinking, Decomposition, Discourse, Main learning outcomes that have been measured
outcomes Efficacy, Formative evaluation, Knowledge, Learning and discussed in the CT research
effectiveness, Mental, Metacognitive, Perception, Problem solving,
Reasoning, Reflection, Student feedback
2. Affective Attitude, Belief, Engagement, Motivation
outcomes
3. Psychomotor Acquisition, Communication, Efficiency, ICT skill, Social,
outcomes Programming skill

The Profile of International Publication Patterns which is over 73% of the total publications from 2006 to
in CT Research 2018. Therefore, the ten countries were labeled as the top
10 productive countries in CT research, namely the USA,
The Most Productive Countries Taiwan, Spain, Turkey, UK, China, Greece, Australia, the
Netherlands, and Canada.
Table 3 lists the distribution of CT research with the The result indicates that over one third of the publica-
number and percentage of their publications, showing tions were from the USA (n = 255, 35.7%); this proportion
publication trends of the most productive countries in CT. can also be applied as the most productive country to
To measure each country’s productivity, the first authors’ period I (n = 87, 35.4%) and period II (n = 168, 35.8%).
country affiliation was counted because the first author has We also found that the number of publications from the
the major contribution to the research. If the first author has USA almost doubled from period I to period II. Countries
multiple affiliations, the main affiliation was manually such as the USA, Spain, the UK, China, and the Nether-
checked and coded. As a result, a total of 52 countries were lands were found to have increasing publication trends in
found to be listed in CT publications. We also found that both numbers and percentage of publications. The second
the first 10 countries have contributed over 535 articles, most publications were from Taiwan (n = 55, 7.7%).

Table 3 The most productive countries on CT publications (2006–2012 vs. 2013–2018)


# Country All years 2006–2018 (n = 715) Period I 2006–2012 (n = 246) Period II 2013–2018 (n = 469) Trend from the
period I to period II
Numbers of Percentage of all Numbers of Percentage of Numbers of Percentage of In In
publications years (%) publications period I (%) publications period II (%) numbers percentage

1 USA 255 35.7 87 35.4 168 35.8 : :


2 Taiwan 55 7.7 26 10.6 29 6.2 : ;
3 Spain 49 6.9 12 4.9 37 7.9 : :
4 Turkey 43 6.0 17 6.9 26 5.5 : ;
5 UK 33 4.6 11 4.5 22 4.7 : :
6 China 29 4.1 9 3.7 20 4.3 : :
7 Greece 23 3.2 10 4.1 13 2.8 : ;
8 Australia 19 2.7 8 3.3 11 2.3 : ;
9 Netherlands 15 2.1 5 2.0 10 2.1 : :
10 Canada 14 2.0 5 2.0 9 1.9 : ;

123
14 K.-Y. Tang et al.

Table 4 Top 20 active authors in CT publications (2006–2012 vs. 2013–2018)


# Authors All year Period I 2006–2012 Period II 2013–2018 First-authored articles Country Cites Active year

Number of articles (h-index)


1 Kafai, Yasmin B. 9 (6) 2 (2) 7 (5) 4 USA 91 2008–2017
2 Denning, Peter J. 8 (4) 5 (3) 3 (2) 8 USA 19 2007–2017
3 Bers, Marina U. 7 (4) 1 (1) 6 (3) 1 USA 20 2012–2018
4 Hwang, Gwo-Jen 7 (4) 2 (2) 5 (2) 1 Taiwan 9 2006–2017
5 Cooper, Stephen 5 (5) 2 (2) 3 (3) 2 USA 9 2009–2015
6 Chen, Gwo-Dong 5 (5) 5 (5) – 2 Taiwan 6 2008–2012
7 Hwang, Wu-Yuin 5 (4) 1 (1) 4 (3) 2 Taiwan 11 2012–2017
8 Magana, Alejandra J. 5 (4) – 5 (4) 3 USA 4 2014–2018
9 Dagiene, Valentina 5 (3) – 5 (3) 3 Lithuania 7 2016–2017
10 Tsai, Chia-Wen 5 (3) 3 (3) 2 (1) 3 Taiwan 3 2008–2017
11 Lau, Wilfred W. F. 4 (4) 3 (3) 1 (1) 4 China 9 2009–2014
12 Chang, Chih-Kai 4 (4) 3 (3) 1 (1) 2 Taiwan 7 2008–2014
13 Ellis, Nick C. 4 (4) 4 (4) – 3 USA 5 2009–2009
14 Burke, Quinn 4 (3) – 4 (3) 2 USA 10 2013–2016
15 Berland, Matthew 4 (3) – 4 (3) 3 USA 9 2013–2015
16 Sentance, Sue 4 (3) – 4 (3) 2 UK 0 2014–2018
17 Roman-Gonzalez, M. 4 (2) – 4 (2) 2 Spain 14 2015–2018
18 Hooshyar, Danial 4 (2) – 4 (2) 4 Malaysia 1 2015–2018
19 Weintrop, David 4 (1) – 4 (1) 2 USA 24 2016–2018
20 Wing, Jeannette M. 2 (2) 2 (2) – 2 USA 385 2006–2008

However, even though the number of publications The remainders were from Lithuania (Valentina Dagiene),
increased from period I (n = 26, 10.6%) to period II China (Wilfred Lau), the United Kingdom (Sue Sentance),
(n = 29, 6.2%), the percentage in period II decreased. Spain (M. Roman-Gonzalez), and Malaysia (Danial
Similar results were found for Turkey, Greece, Australia, Hooshyar). In other words, over half of the active authors
and Canada (downtrend signs in percentage). This reveals were affiliated with the US and the western universities/
the fact that the overall number of international publica- research institutions; but almost the other half of the
tions on CT has increased, and the global competition in authors were from the Asia–Pacific region.
CT research has become more intense, especially in recent Taking results of CT publication patterns altogether, we
years. found that the dominant productivity and impact in CT
research were from the western countries, especially the
The Most Active Authors USA. However, some growth trends from non-western
countries were also noteworthy. For example, three of the
The analysis of the most active authors helps identify top ten countries are Taiwan, Turkey, and China (Table 3),
active researchers with frequent publications that continue and seven out of top 20 active authors were also from non-
contributing new findings for benefitting the CT research western countries (Taiwan, China, and Malaysia) (Table 4).
community. Table 4 lists top 20 active CT researchers We also observed that eight researchers from various
worldwide. Among a total of 1652 researchers, Yasmin regions who were not published in period I became active
Kafai was the most active author who had published nine authors in period II, showing that CT research has attracted
articles with h-index 6 and 91 citations from 2008 to 2017. many international researchers.
The results also show an increasing trend of her continued
publications from period I (n = 2) to period II (n = 5). The The Most Influential Research Articles
second-most active author was Peter Denning with eight
first-authored articles, h-index 4 and 19 citations, followed To more specifically discover the publication trend, the
by Marina Bers, with seven articles with h-index 4 and 20 most influential research articles with highly cited times
citations. Overall, a total of 10 active authors were from the were reviewed. Note that only the articles cited by other
USA, and five were from Taiwan (led by Gwo-Jen Hwang). journal articles in SSCI/SCI were counted as citations.

123
A Content Analysis of Computational Thinking Research: An International Publication Trends and… 15

Overall, the top three highly cited articles have received at educators to encourage and inspire students to think like
least 201 citations in total and 20.1 times on average (per computer scientists.
year). The most highly cited article is Wing’s (2006)
original article titled ‘‘computational thinking,’’ advocating The Main Characteristics of CT Research
the importance and practices of CT for every learner (302
total citations, 23.2 cites per year). Followed by Papaster- Based on the keyword-derived typology (Table 2), three
giou’s (2009) empirical study in examining students’ main characteristics of CT research were analyzed,
learning effectiveness and motivations in learning the including background settings, domain-specific factors, and
computer memory concept via computer games. He con- learning outcomes. Each dimension was reported with the
cluded that that the gaming approach was effective in sub-categorizations and counts of each occurring keyword.
promoting students’ knowledge of computer memory Note that duplications of each keyword from individual
concepts and more motivational than the non-gaming studies were carefully filtered; the number of counts rep-
approach (286 total citations, 28.6 cites per year). The third resents the frequencies of such wording being used in the
one is Resnick et al’s (2009) article titled ‘‘Scratch: pro- number of articles.
gramming for all,’’ introducing a game-based programming
tool that facilitating individuals to access coding activities Background Settings
(201 total citations, 20.1 cites per year).
Recently, Denner et al. (2012) conducted experiments Figure 2 shows the results of the background settings,
among middle school female students by using a visual which were further divided into participants and subjects.
tool, called Statecast Creator, to program a computer game. In the first category of participants, secondary students
They found that game construction involving both design (n = 118) were found as the major research target in CT
and programming activities can enhance students’ learning research, followed by those from higher education
of computer science concepts. Weintrop et al.’s (2016) (n = 109) and elementary schools (n = 72), whereas tea-
position paper articulated the benefits of including com- cher (n = 27) and kindergarten (n = 10) participants were
putational thinking in mathematics and science classrooms. the two least frequent. Next, the results reveal that sec-
They point out that embedding CT in a learning curriculum ondary students were also the most discussed participants
can establish a reciprocal relationship between CT and in period II (30 88), followed by those from higher edu-
learning in mathematics and science, and bring math and cation (42 65), even though they were found the most in
science education closer to current professional practice. In period I. Teachers, however, were still found to be the least
addition, these articles have integrated various game-pro- (6 21). This result reveals that there is a lack of teacher
gramming tools into formal curricula. This indicates that training for CT. This indicates that fostering CT is still a
game design can facilitate CT cultivation, enable learners challenge due to only a few teachers being trained with the
to solve problems in real life, and empower them the ability knowledge and skills to integrate CT into course curricula.
to apply educational knowledge into the practical world. Similarly, Mouza et al. (2017) urged that CT skills must be
From these highly cited papers, we found that the idea of taught in teacher education curricula to foster deeper
computational thinking does help computer developers and knowledge and skills so that they can graduate well

Fig. 2 Counts and trends of Secondary (118) 30 88


1. Participants

background settings on CT Higher education (109) 42 65


research from 2006 to 2018.
Elementary (72) 16 56
(Number) denotes total counts
for each subcategory from 2006 Teacher* (27) 6 21
to 2018. *Including in-service Kindergarten (10) 3 7
teacher, school teacher, and Computer science (78) 32 46
teacher training. **Including Mathematics (65) 17 48
Linguistic and Morphology Engineering (59) 12 47
Arts (36) 12 24
2. Subjects

Robotics (25) 4 21
Languages**(19) 8 11
Biology (11) 2 9
Physics (11) 4 7
Chemistry (7) 16
Period I (2006-2012) Period II (2013-2018)
Music (7) 34
0 20 40 60 80 100 120 140

123
16 K.-Y. Tang et al.

Game (77) 17 60
Peer/ Collaboration (68) 23 45
Simulation (37) 9 28
Inquiry (13) 3 10

1. Pedagogy
Self-regulated learning (11) 83
Problem-based learning (10) 55
Microworld (7) 43
Pair programming (6) 33
APOS* (5) 23
Situated (4) 31
Constructivism (4) 04
Environm

Web-based/Online (160) 68 92
ent

Face-to-face** (94)
2.

34 60
Blended/ hybrid (13) 8 5
Scratch (44) 4 40
Java (14) 9 5
3.programming languages/ Tools

C/ C++ (11) 47
Robots (11) 29
Lego (6) 15
R language (5) 32
Alice (5) 14
Python (4) 04
JavaScript (2) 02
Kodu (2) 02
Code.org (1) 01 Period I (2006-2012) Period II (2013-2018)
Augmented Reality (1) 01
0 20 40 60 80 100 120 140 160 180

Fig. 3 Counts and trends of domain-specific factors on CT research from 2006–2018. *APOS = APOS (Action, Process, Object, Schema) theory
(Dubinsky and McDonald, 2001). **Including Face-to-face and laboratory. ;: downward trend sign

prepared. In addition, it is also noteworthy that the pedagogical methods dominate the CT research in both
youngest students from kindergarten were found to be period I and II with upward trends, especially games (17
selected as participants in CT research from period I 60). Other approaches, such as inquiry (n = 13), problem-
(n = 3) to period II (n = 7). All of the analyzed participants based learning (n = 10), constructivism (n = 4), APOS
showed uptrend signs for the two research periods. (n = 5), and pair programming (n = 6) were found to be
In the second category of subjects, as expected, Com- equal or even more in period II than period I. These used
puter Science (n = 78) was the main subject in CT pedagogies reflect the fact that the development of CT is
research. This was followed by Mathematics (n = 65), needed to create task-based learning environments to
Engineering (n = 59), and Robotics courses (n = 25), enhance motivation and cognitive outcomes (Belland et al.
indicating that CT was highly discussed in STEM educa- 2013). This also implies that instructors may need to
tion. Other subjects were also found across natural science, construct adaptive learning environments to scaffold stu-
such as Biology (n = 36), Physics (n = 36), and Chemistry dents to take more control in CT learning (Drexler 2010).
(n = 7); and the humanity and arts fields, including Arts Next, regarding learning environments, web-based/
(n = 36), Languages (n = 19), and Music (n = 7). All online (n = 160) was the most frequent setting adopted in
subjects also showed an uptrend from period I to period II. CT research, followed by face-to-face environments
(n = 94). Both showed an uptrend in the two periods.
Domain-Specific Factors However, the results of blended/ hybrid (8 5) which
specifies class settings with mixed environments was found
Figure 3 shows the results of the domain-specific factors, to be less or was even gradually being replaced by today’s
including pedagogy, environment, and programming lan- mobile learning era. It is also noticeable that the differ-
guage/ tools. First, game (n = 77) was the most frequently ences between online and face-to-face decreased in period
adopted pedagogy in CT research, followed by peer/ col- II. CT enrichment activities can be taught with or without
laboration (n = 68), and simulation (n = 37). These the help of technologies. For example, recently educators

123
A Content Analysis of Computational Thinking Research: An International Publication Trends and… 17

Knowledge (151) 53 98
Mental (133) 41 92
Academic (39) 19 20
Decomposition (48) 17 32
Abstraction (54) 13 41

1. cognitive domain
Student feedback (37) 12 25
Critical (44) 6 37
Perception (32) 13 19
Efficacy (36) 14 22
Reasoning (28) 5 23
Reflection (24) 5 19
Problem-solving (9) 4 5
Discourse (8) 4 4
Metacognitive (6) 24 Period I (2006-2012) Period II (2013-2018)
Learning effectiveness (3) 30
Formative evaluation (2) 20
Motivation (55) 18 37
3. psychomotor 2.affective
domain

Attitude (43) 13 30
Engagement (30) 3 27
Belief (18) 9 9
Social (87) 22 65
Communication (49) 23 26
domain

Programming skill (28) 8 20


Acquisition (20) 8 12
Efficiency (11) 8 3
ICT skill (3) 30
0 20 40 60 80 100 120 140 160

Fig. 4 Counts and trends of learning outcomes on CT research from 2006–2018

have adopted board games for younger learners or adult measured the learning outcomes as decomposition
beginners to develop CT skills, including basic coding for (n = 48), abstraction (n = 54), student feedback (n = 37),
8–9-year-old learners (Tsarava et al. 2018) and collabora- critical thinking (n = 44), efficacy (n = 36), reasoning
tion, communication, and problem solving for groups of (n = 28), reflection (n = 24), discourse (n = 8), and
college-age novice players (Berland and Lee 2011). These metacognition (n = 6). Nonetheless, academic (n = 39),
changes in the learning environment leave a clue for edu- formative evaluation (n = 3), and learning effectiveness
cators to be aware of the importance of learners’ CT (n = 2) were found but appeared less in period II. These
cultivation. cognitive findings reflect the importance of CT enhance-
Lastly, the use of programming languages/tools was also ment, as it is a heuristic higher-order thinking skill devel-
discussed. The results show that Scratch (n = 44) was opment for learners.
heavily used in CT research with a great jump from period Next, in the affective domain, participants’ motivation
I (n = 4) to period II (n = 40), and that Java (n = 14) (n = 55) was found to be measured most often, while
occurred most in the earlier period. Note that the use of attitude (n = 43), engagement (n = 30), and belief (n = 18)
Java and R language was also found to decrease; however, were considered as learning goals in the more recent CT
robots, Lego, and Alice were found to be increasingly research. Among them, engagement was found to be less
studied in recent years. Those that were found active in addressed in period I, but much more often in period II (3
recent years were tools more user-friendly, or game-based, 27). This result can be viewed as evidence linking with
including Python, JavaScript, Kodu, and Code.org. One Wing’s (2006) argument. ‘‘Computational thinking is
recent application of augmented reality in CT research was reformulating a seemingly difficult problem into one we
also found in this analysis (Teng et al. 2018), showing know how to solve, perhaps by reduction, embedding,
some needs for visual-based features for programming transformation, or simulation’’ (Wing 2006, p. 1). Thus, CT
learning for the younger generation. education helps develop individuals’ higher-order thinking
skills to cope with the fast-paced society and the mixed
Learning Outcomes reality environments. Facilitating adaptive and engaging
environments to nourish intrinsic motivation for CT
As shown in Fig. 4, the results of the learning outcomes developments for learners of all ages is the key.
were divided into the cognitive, affective, and psychomotor Lastly, in the psychomotor domain, social (n = 87) and
domains. In the cognitive domain, participants’ knowledge communication (n = 49) skills were found to be frequently
(n = 151) and mental understanding (n = 133) were most evaluated in CT research, where social skills were more
frequently assessed. While these two words show general often addressed in period II (22 65). In addition, skills such
meanings for cognitive assessments, some researchers as programming (n = 28), acquisition (n = 20), efficiency

123
18 K.-Y. Tang et al.

(n = 11), and ICT (n = 3) were also found. However, this study is the keywords, especially from article’s
efficiency and ICT skills may already be a basic ability for abstract. In general, the abstract provided at most 300
online-generation students after two decades of broadband words. Moreover, writing style varied among researchers:
network development. some researchers prefer concise wording; others may tend
to give details of the research backgrounds and results, and
even according to the rules of each journal. In this study,
Conclusions and Limitations about 56% of articles have reported their research results in
the abstract; it is difficult to comment which outcomes
This study conducted keyword and content analyses on CT were superficial and which were more critical by merely
research in the recent 13 years. All in all, our result showed using keyword analysis. It is the same for the relationships
that research publications across the two analyzed periods between the CT use and curricular topics, as well as the
have greatly increased and have spread around the world in relationship between CT concepts/practices and higher-
recent years. In terms of productivity, the USA dominated order skills. As a complementary research method, meta-
the CT publications in both periods. However, the rising analysis and extensive content analysis are suggested for
impacts from Asia and Europe are noteworthy, with several further examination of the above relationships. Last, to
active researchers from these countries being identified. present the latest research status of CT, the timespan set-
We also have provided a typology of CT research that ting is not a complete year (until August, 2018). Future
was based on the results of the keyword analysis and studies can gather data from a complete research year to
content analysis. Several implications for the future continue the observations of CT development. Bearing the
research were concluded. First, the CT-infused courses limitations in mind, this keyword analysis of CT publica-
were most often conducted in STEM education, and some tions has provided a fresh attempt to demonstrate a con-
humanity and arts domains. This finding highlights that the ceptual and typological overview with bibliographic
philosophy of CT can be directly applied to natural and evidence to quickly follow the research status and key
social sciences. Second, pedagogy using games, peer/col- elements of CT literature.
laboration approaches, and task-based learning environ-
ments were found. This higher-order thinking cultivation Acknowledgements The authors are grateful to the comments from
two anonymous reviewers. This work was financially supported by
environment can foster students’ computational practices the ‘‘Institute for Research Excellence in Learning Sciences’’ of
with information processing, scaffolding, and reflection National Taiwan Normal University (NTNU) from The Featured
activities. Moreover, Scratch, Lego, Alice, and Python Areas Research Center Program within the framework of the Higher
were found to be emerging programming language/ tools, Education Sprout Project by the Ministry of Education (MOE) in
Taiwan. Partly financial support from the Ministry of Science and
highlighting the use of game-programming tools as a Technology, Taiwan, under grant number MOST 107-2511-H-130-
means of helping learners develop the skills to face real-life 003 was also acknowledged.
problems.
This study has contributed a research typology from
keyword and content analyses to understanding the litera- References
ture structure of computational thinking research, and has
revealed some changes in three major characteristics, Belland, B. R., Kim, C., & Hannafin, M. J. (2013). A framework for
designing scaffolds that improve motivation and cognition.
namely background settings, domain-specific factors, and Educational Psychologist, 48(4), 243–270.
learning outcome domains. Nevertheless, all methodology Berland, M., & Lee, V. R. (2011). Collaborative strategic board
has their own research limitations (Nerur et al. 2008). In games as a site for distributed computational thinking. Interna-
this study, the first limitation is the completeness issue of tional Journal of Game-Based Learning, 1(2), 65–81.
Bloom, B. S., Engelhart, M. D., Furst, E. J., Hill, W. H., & Krathwohl,
keyword-searching strategies. Even though we used 13 D. R. (1956). Taxonomy of educational objectives: The classi-
keywords derived from four influential papers, the cover- fication of educational goals. Handbook I: Cognitive domain.
age may not be complete. For example, the works related to New York: David McKay Company.
‘‘computational action’’ and ‘‘system thinking’’ used in the Buitrago Flórez, F., Casallas, R., Hernández, M., Reyes, A., Restrepo,
S., & Danies, G. (2017). Changing a generation’s way of
educational simulation area might have been omitted in our thinking: Teaching computational thinking through program-
analysis. However, following the same criteria in this paper ming. Review of Educational Research, 87(4), 834–860.
by searching the category of education and educational Cook, D. A., Beckman, T. J., & Bordage, G. (2007). A systematic
research from the WoS, less than 20 new articles were review of titles and abstracts of experimental studies in medical
education: Many informative elements missing. Medical Edu-
obtained using the above two keywords. This indicates a cation, 41(11), 1074–1081.
very limited impact on our results. For future analysis, Denner, J., Werner, L., & Ortiz, E. (2012). Computer games created
researchers may add more keywords and search a broader by middle school girls: Can they be used to measure
category to enlarge the data. Second, the unit of analysis in

123
A Content Analysis of Computational Thinking Research: An International Publication Trends and… 19

understanding of computer science concepts. Computers & citation analysis. Strategic Management Journal, 29(3),
Education, 58(1), 240–249. 319–336.
Drexler, W. (2010). The networked student model for construction of Papastergiou, M. (2009). Digital game-based learning in high school
personal learning environments: Balancing teacher control and computer science education: Impact on educational effectiveness
student autonomy. Australasian Journal of Educational Tech- and student motivation. Computers & Education, 52(1), 1–12.
nology. https://doi.org/10.14742/ajet.1081. Park, J., Wu, B., Morrison, A. M., Shen, Y., Cong, L., & Li, M.
Dubinsky, E., & McDonald, M. A. (2001). APOS: A constructivist (2016). The tourism system research categorization framework.
theory of learning in undergraduate mathematics education Asia Pacific Journal of Tourism Research, 21(9), 968–1000.
research. In The teaching and learning of mathematics at Resnick, M., Maloney, J., Monroy-Hernández, A., Rusk, N.,
university level (pp. 275–282). Dordrecht: Springer. Eastmond, E., Brennan, K., et al. (2009). Scratch: Programming
Dyrbye, L. N., Harris, I., & Rohren, C. H. (2007). Early clinical for all. Communications of the ACM, 52(11), 60–67.
experiences from students’ perspectives: A qualitative study of Shailaja, J., & Sridaran, R. (2015). Computational thinking the
narratives. Academic Medicine, 82(10), 979–988. intellectual thinking for the 21st century. International Journal
Glaser, B. G., & Strauss, A. L. (1967). The discovery of grounded of Advanced Networking & Applications, 7, 39–46.
theory: Strategies for qualitative theory. New Brunswick: Aldine Shute, V. J., Sun, C., & Asbell-Clarke, J. (2017). Demystifying
Transaction. computational thinking. Educational Research Review, 22,
Grover, S., & Pea, R. (2013). Computational thinking in K–12: A 142–158.
review of the state of the field. Educational Researcher, 42(1), Teng, C. H., Chen, J. Y., & Chen, Z. H. (2018). Impact of augmented
38–43. reality on programming language learning: Efficiency and
Hsiao, C. H., Tang, K. Y., & Liu, J. S. (2015). Citation-based analysis perception. Journal of Educational Computing Research,
of literature: A case study of technology acceptance research. 56(2), 254–271.
Scientometrics, 105(2), 1091–1110. Tsarava, K., Moeller, K., & Ninaus, M. (2018). Training Computa-
Kim, H., Choi, H., Han, J., & So, H. J. (2012). Enhancing teachers’ tional Thinking through board games: The case of Crabs &
ICT capacity for the 21st century learning environment: Three Turtles. International Journal of Serious Games, 5(2), 25–44.
cases of teacher education in Korea. Australasian Journal of Voogt, J., Fisser, P., Good, J., Mishra, P., & Yadav, A. (2015).
Educational Technology, 28(6), 965–982. Computational thinking in compulsory education: Towards an
Lin, T. J., Lin, T. C., Potvin, P., & Tsai, C. C. (forthcoming). agenda for research and practice. Education and Information
Research trends in science education from 2013 to 2017: A Technologies, 20(4), 715–728.
systematic content analysis of publications in selected journals. Weintrop, D., Beheshti, E., Horn, M., Orton, K., Jona, K., Trouille, L.,
International Journal of Science Education. et al. (2016). Defining computational thinking for mathematics
https://doi.org/10.1080/09500693.2018.1550274. and science classrooms. Journal of Science Education and
Liu, J. S., Ho, M. H. C., & Lu, L. Y. (2017). Recent themes in social Technology, 25(1), 127–147.
networking service research. PLoS ONE, 12(1), e0170293. Wing, J. M. (2006). Computational thinking. Communications of the
Lye, S. Y., & Koh, J. H. L. (2014). Review on teaching and learning ACM, 49(3), 33–35.
of computational thinking through programming: What is next Wing, J. M. (2008). Computational thinking and thinking about
for K-12? Computers in Human Behavior, 41, 51–61. computing. Philosophical Transactions of the Royal Society of
Maida, V., & Cheon, P. M. (2014). Prognosis: The ‘‘missing link’’ London A: Mathematical, Physical and Engineering Sciences,
within the CanMEDS competency framework. BMC Medical 366(1881), 3717–3725.
Education, 14, 93. https://doi.org/10.1186/1472-6920-14-93. Zawacki-Richter, O., & Latchem, C. (2018). Exploring four decades
Mouza, C., Yang, H., Pan, Y. C., Ozden, S. Y., & Pollock, L. (2017). of research in Computers & Education. Computers & Education,
Resetting educational technology coursework for pre-service 122, 136–152.
teachers: A computational thinking approach to the development
of technological pedagogical content knowledge (TPACK). Publisher’s Note Springer Nature remains neutral with regard to
Australasian Journal of Educational Technology, 33(3), 61–76. jurisdictional claims in published maps and institutional affiliations.
Nerur, S. P., Rasheed, A. A., & Natarajan, V. (2008). The intellectual
structure of the strategic management field: An author co-

123

You might also like