Computational thinking: Visible in the classroom but invisible in the curriculum

In our technology-intensive world, computing and programmed technological solutions have gained in importance, and their influence on curriculum, teaching, and learning has been substantial worldwide. Sweden, along with many countries, has integrated programming into the compulsory school curriculum as an integrated part of the teaching of Mathematics and Technology. In addition to a focus on programming, the new curricula also place significant emphasis on digital skills and on enhancing awareness of how the digitalisation of society affects us. Programming is described as a digital competence and computational thinking (CT) as important knowledge through which to facilitate learning and understanding of programming. Thus, it seems that CT, as seen in the Swedish context, should relate to both programming and digital competence. In this article, the aim is to examine the presence of CT in Swedish research literature and as a part of the discourse around the development of the Swedish curriculum. A content analysis of the curriculum and a thematic analysis of research publications show that CT is not well integrated into Sweden’s educational system. However, CT-related activities are found in several subjects and research about CT, and programming is thriving. Requirements for the design of complex systems where understandings of humans and technology are equally important put new demands on education. Meeting these demands in education can be a challenge, but one subject in the Swedish curriculum seems to be suitable for the task, the technology subject. We conclude that the subject technology should be revised to include a greater focus on creativity regarding CT and the construction of computational technological artefacts.


Introduction
Computational thinking (CT) is a widely debated and elusive concept (Denning & Tedre, 2019;Pears et al., 2021).It is also doubtful whether today's focus on programming and coding gives students the right knowledge to live in an even more digitalised world in which artificial intelligence and machine learning systems are likely to dominate (Pears et al., 2021).This article examines the concept's position in technology education, taking Sweden as an example.Our primary data sources are the Swedish curriculum (Skolverket, 2018a(Skolverket, , 2018b) ) for compulsory school and research on CT related to Swedish education (Appendix 1).
It is difficult to discuss CT and education without considering Seymour Papert.Piaget's knowledgebased learning (Papert, 1980(Papert, /2020) influenced Papert to develop two dimensions of Piaget's theory in relation to how working with computers can support the development of, [1] intellectual structures in the child and [2] "the design of learning environments that are resonant with them" (1980/2020, p.181).The environment that Papert described was designed to support children's learning in mathematics and physics by "doing science" (1980/2020, p. 107).
The artificial world was important to Papert and things (computational artefacts) occupied the learning environment he created.It was in the interaction and communication with these "things" that learning took place (1980/2020).Papert described computers and robots as learning aids and the child communicated with these devices using an artificial turtle and the language LOGO (1980/2020).When the child tried an idea, the turtle showed the result as a drawing on the computer screen or with the help of a robot drawing the results on paper.It was important to Papert that the child was in control over the interaction with the technical artefacts and he argued, " . . .my central focus is not on the machine but on the mind . . ." (1980/2020, p. 8).
Papert described the potential for learning through child-machine interaction already in the 1980s, but robots and computers became everyday educational tools much later.Teachers lacked knowledge in computer science and CT.Instead, teaching focused on skills with digital tools, such as word processing software and desktop publishing (Denning & Tedre, 2019).Some decades after Papert, Jeanette Wing attracted considerable attention with her article "Computational Thinking" (Wing, 2006).She described a vision of a generalised application of computer science principles, arguing that a way of thinking using concepts and approaches from computer science will be a central competency area for future citizens.She argued that computational processes, with or without computers, will be central in understandings of the modern world.Both Papert and Wing emphasised the human aspect of the interaction between people and computers.Papert emphasised the exploration of ideas and Wing argued for CT as a fundamental skill (Papert, 1980(Papert, /2020;;Wing, 2006).Although Papert and Wing are influential, findings in this article demonstrate that there is a productive field of research active in research related to CT and teaching.
Much attention in society, however, has been on the skills of instructing the machine, on programming (Pears et al., 2021;Wanngård et al., 2015).Sweden, along with many countries, responded to political and societal pressure to address the need for these types of skills by integrating programming into the compulsory school curriculum (Wanngård et al., 2015).A revised curriculum was published in 2018 (Skolverket, 2018a(Skolverket, , 2018b) ) with the aim of strengthening students' digital competence (Utbildningsdepartementet, 2017).The same year the Swedish National Agency for Education also published a research overview, disseminating knowledge on how digitalisation affects education (Skolverket, 2018c).The report described programming as a digital competence and CT as important for gaining an understanding of programming (Skolverket, 2018c).Concerning these directives, CT appears to be important and related to both programming and digital competence.The revision of the c.Which perspectives on CT are found in the publications?
The content analysis results indicate that the curriculum provides weak support for CT.On the other hand, the thematic analysis presents several perspectives of CT in the research literature.Despite the weak guidance from policy documents, CT content is present in school practice.

Method
To investigate how CT is integrated into the Swedish curriculum and in school practice, we analyse two primary data sources, the Swedish school curriculum for the compulsory school revised 2018 and research literature published in the period 2015 to 2021.The analysis objects are hence documents.According to Rapley (2007), documents are not just containers with objective information but tell about the specific context in which they are used."We never just somehow neutrally or abstractly engage with documents, they are always engaged with in [sic] a specific local context; as such, they are always read or used in a specific way, to do specific work."(Rapley, 2007, p. 97).In line with Rapley's (2007) argument, we consider our data sources to reflect conditions for teaching and learning.The curriculum is analysed with content analysis and the research literature with thematic analysis.

Content analysis
Communicative material is the basis for content analysis.Examples of communicative material are transcripts of interviews, answers to open-ended survey questions and electronic written communication (White & Marsh, 2006).Quantitative content analysis and qualitative content analysis originate from different research traditions.Quantitative content analysis stems from a positivist tradition and uses statistical methods to test and present hypotheses.Qualitative content analysis stems from a humanistic research tradition and answers research questions by examining and identifying patterns in texts (White & Marsh, 2006).
Summative content analysis is a qualitative content analysis despite its procedure of counting keywords to explore the use of certain words in a text.The word count reveals patterns in the text and opens up for interpretation (Hsieh & Shannon, 2005).Prior (2003) states that words, as part of a text, are references to the social context described by the text.In this sense, reference counting in content analysis is a powerful tool for revealing patterns and providing information about the context in which the text is used.In this study is summative content analysis used to answer the research question: a.What conditions provide the curriculum for CT to appear as a prioritised subject matter?However, the number of references in a content analysis will not provide a complete understanding of the context (Prior, 2003).The use of a text needs to be studied to fully understand the relationship between text and context.Therefore, the content analysis in this study does not provide answers about how teachers interpret and use the curriculum.Rather, the analysis shows what conditions exist for teaching based on the support the curriculum provides.
Two documents in PDF format were analysed.Firstly, the curriculum for the compulsory school, preschool class, and school-age educare (Skolverket, 2018b).This curriculum was in Swedish and revised in 2018.Second, the same curriculum in English (Skolverket, 2018a).The tools used in the analysis process were Acrobat Reader, Excel, and NVivo.
The compulsory school in Sweden is preschool class (one year) and grades 1 to 9. The curriculum consists of three parts: fundamental values and tasks of the school, overall goals and guidelines, and syllabi.In our analysis, the Swedish words digital (digital), programmering (programming), and datalogi/sk/t (computational) were counted as references.Combinations or variations of these words in the curriculum were included in the analysis.Each word combination was counted as one reference.For example, datalogiskt tänkande (computational thinking) is a combination of datalogi/sk/t (computational) and tänkande (thinking) and counted as one reference, see Figure 2.

Rapid literature review
A rapid literature review was used to find research related to teaching and learning of CT in Swedish schools.The publications, the results of the rapid literature review, were later analysed with thematic analysis.
The process for our review follows the characteristics of a rapid literature review in the sense that the selection process was systematic, had a narrow focus, and was completed within a limited time (approximately 12 months, not full-time) (Denscombe, 2016;Folkhälsomyndigheten, 2017;Ganann et al., 2010).Rapid literature reviews can be extensive and form the basis for authorities' decisions (Folkhälsomyndigheten, 2017) or be the ground for analysis in small-scale qualitative research studies (Denscombe, 2016).The rapid literature review in this study was used in a small-scale qualitative research project.Compared to a systematic review, a rapid review may miss some perspectives since it is not as comprehensive as a systematic review and may not cover everything.Therefore, we do not claim to cover all research, but only that which emerged through the search procedure shown in Figure 1.
In order to find relevant research, we used the databases SCOPUS, ERIC, and Web of Science.Different combinations of keywords were tested.The best results were generated when the following terms were combined: computational thinking; school; education; Sweden; and programming.The term programming was added to broaden the search, while the term school was added to exclude studies related to higher education.The keyword combinations and examples of search strings are shown in Table 1.The search in four databases resulted in four parallel search and selection procedures (Figure 1).The following selection procedure is used:  After screening and selection, 24 research articles and papers were identified (Figure 1; Appendix 1).Three publications focus on preschool, one on upper secondary school and 20 refer to primary or grades K-12.Ten have been published prior to, and 14 after, the publication of the revised Swedish curriculum in 2018.

Thematic analysis
The thematic analysis is a qualitative method for identifying patterns within data.It is a flexible method independent of theory and therefore it is important for the researcher to have an active role and account for the strategies used in the analysis (Braun & Clarke, 2006).One way to start the analysis is the theoretical approach.Data for a particular analysis is selected from the corpus, that is, the entire research material.This data set is then searched for patterns and the patterns are coded in relation to the specific research question.The research question can also be developed through coding, an inductive approach (Braun & Clarke, 2006).Regardless of the chosen approach, similar codes are merged into themes.Ryan and Bernard (2003) respond to descriptions of thematic analysis where themes appear to emerge without explanation and emphasise the active role of the researcher in the search for knowledge and theoretical understanding.The researcher should thus be consistent when defining themes and report on the decisions made during the analysis.Braun and Clarke (2006) distinguish between semantic or latent levels of identification.Semantic identification is to identify a theme through descriptions from the data and thereby give it meaning on a superficial, explicit level.The interpretation is later built upon and summarises the descriptions.Latent identification of themes, on the other hand, goes beyond the semantic content of the data and examines underlying ideas.
Reading the corpus and generating alternate codes and themes, the procedure continues and ultimately results in an interpretation that summarises the broad meaning of the data (Braun & Clarke, 2006;Ryan & Bernard 2003).In this study, we use a thematic analysis to answer the following research questions: b. Which frameworks are used to define CT in the publications?c.Which perspectives on CT are found in the publications?
Initially, the data set was analysed to find definitions of CT.When the data provided opportunities for further analysis, research questions (b) and (c) were developed.In this way, the analysis began theoretically, but an inductive approach was gradually developed.The analysis was close to the text and themes were identified at a semantic level.Table 2 shows examples of data extracts generated into codes and mapped to a theme.The first excerpt describing the importance of Papert and Wing was coded separately into two codes, Papert and Wing (2006).The second excerpt refers to a framework by Brennan and Resnick.This extract was coded to Brennan and Resnick (2012).All three codes were mapped to the theme Frameworks (Table 2).
The entire analysis followed the following steps in line with previous research (Braun & Clarke, 2006;Ryan & Bernard 2003).
Selection of 24 publications based on the rapid literature review (Appendix 1) -The corpus was imported into the analysis software, NVivo.Excel was used as a supplementary tool.-Descriptions in the text were coded at the semantic level.
-Codes were identified through an initial theoretical approach.
-An inductive approach generated research questions (b) and (c) -Reading of corpus and individual publications was alternated.
-Codes were compiled into themes (Table 2) -Codes and themes were generated until the themes provided clear meaning to the data (Braun & Clarke, 2006).

Frameworks
Australasian Journal of Technology Education, Special Issue: Technology Education on the Edge, Vol. 9, 2023 7 computational practices, and computational perspectives as its three key dimensions.(Humble et al., 2019, 2.4 Computational thinking, para.2)

Results
The results consist of two parts.First, a summative content analysis of the Swedish curriculum (Skolverket, 2018a(Skolverket, , 2018b)).Second, a thematic analysis of the research publications provided with the rapid literature review.

A summative content analysis of the curriculum
A summative content analysis examined research question (a).This question concerns what conditions the curriculum provides for CT to emerge as a priority subject.
Table 3 shows the presence in the curriculum of the references Computational Thinking, Programming, and Digital (Skolverket, 2018a(Skolverket, , 2018b)).Table 3 indicates the word Yes, the presence of references and the word No, the absence of references.Digital is in all but three syllabi and there are no references at all to Computational Thinking.
Programming references are in the syllabi for three subjects: Civics, Mathematics, and Technology (Table 3).It is clear that references to Digital dominate the curriculum.

School-age educare Yes No No
Figure 2 shows all variants and word combinations included in the references Digital, Programming and Computational Thinking.The difference in the number of references to Digital is markedly higher (178 references) than Programming (14 references).The most common references are Digital tools, followed by Digital media and Digital aid/s.Figure 2 confirms that there are no references to Computational Thinking.Found in almost all syllabi was Digital tools, the most common reference.Digital media, the second common reference, was in about a third of syllabi and in other parts of the curriculum.In comparison, the third most common reference, Digital aid/s, was only in the optional courses, Modern languages and Mother tongue tuition.Based on these differences Digital tools and Digital media was in the analysis suggested to have a greater impact on the curriculum as a whole than Digital aid/s.

A thematic analysis of research literature
The basis for the thematic analysis is the literature selected in the rapid literature review (Appendix 1) and further on are the results presented.The first results are in line with research question (b).This question concerns the frameworks used in the publications to define CT.The following results are in line with research question (c).This question concerns perspectives on CT found in the publications.

Framework used in the research literature
Fifteen frameworks were identified in the publications and mapped to the Theme Frameworks (Appendix 2). Figure 3 shows all codes mapped to this theme, most common were references to Wing (2006) followed by Papert.
Definitions of CT derived from Papert's work were learning by doing, exploring with code; testing and debugging; to construct artefacts from own ideas; and robot programming for children.Definitions of CT as a universal ability useful for all were derived from Wing's work.CT was regarded as a specific thinking or logic derived from computer science regarding problem solving in interaction with information agents e.g., computers.
Furthermore, definitions of CT other than Papert's and Wing's are listed below.
• Barr et al. (2011) stressed that certain attitudes are important when practising CT (referenced in Vinnervik, 2020).• Bers et al. (2014) emphasised practice and argued for that certain mental tools and attitudes are in use when exploring and creating (referenced in Otterborn et al., 2020;Palmer, 2017;Zhang et al., 2020).• Lye and Koh (2014) underlined the relationship between children and the digital world and the need for digital literacy when solving problems with programming (referenced in Bowden, 2019;Kjällander et al., 2021;Palmer, 2017).• Perkovic (2015) (referenced in Mozelius & Öberg, 2017) addressed the understanding of processes and, more specifically, computational processes.• Tedre and Denning (2016) (referenced in Vrieler, 2017) emphasised the development of CT through professional practice.
Australasian Journal of Technology Education, Special Issue: Technology Education on the Edge, Vol. 9, 2023 10 • Shute et al. ( 2017) described a model for CT with six aspects (referenced in Humble et al., 2019Humble et al., , 2020)).• Zhang and Nouri (2019) developed the framework of Brennan and Resnick (2012) with one concept, three practices, and one perspective (referenced in Zhang et al. 2020).

Figure 3. Number of publications referencing each framework.
There was no uniform definition of CT in the research reviewed, although some of the frameworks had overlapping terminology.An all over conclusion is that the legacy of Papert and Wing is clear.It is also clear that CT is a dynamic construct still evolving; for example, Zhang et al. (2019) building on the framework of Brennan and Resnick (2012) (referenced in Zhang et al., 2020).

Perspectives on CT in the research literature
Six themes were identified in the thematic analysis, Themes A-F (Table 4).These themes represent different perspectives of CT found in the publications.Theme (A) complements the findings on frameworks in line with research question (a).The number of definitions of CT in Appendix 3 therefore matches Theme A. Theme (B) describes CT as problem solving and is an additional finding in line with research question (a).
Theme (C) concerns CT and programming.It was hard to find distinctions between CT and programming in the analysis, as the two concepts were intertwined and hard to separate.Furthermore, it is not within the scope of this article to define programming.However, the analysis identified some distinctions.In some publications CT was described as the logic and strategies behind the problemsolving design.On the other hand, other publications described the development of CT abilities through programming activities.
Terms like language, code, instruction, and tools related to programming.For example, arguments about block versus text-based programming as suitable tools for different age groups.
Theme (D) concerns CT as a universal competence useful in the modern digital society.In the publications, general problem-solving ability, understanding of computational processes and products were examples of this competence.Theme (D) also complements findings in line with research question (a).Themes (E) and (F) focus on educational issues.Theme (E) contains arguments concerning both the revised curriculum from 2018 and the implementation of CT and programming.Other topics discussed were assessment of CT skills, progression of CT skills, and CT for young children.
In Theme (F), CT in school subjects, is CT related to school subjects and the theme is further developed in Table 5.

Table 4. Themes Derived from the Thematic Analysis
Perspectives on computational thinking (CT) derived from the reviewed publications.
Australasian Journal of Technology Education, Special Issue: Technology Education on the Edge, Vol. 9, 2023 12 Technology CT in Technology is linked to the control of objects, for example robots.Simulations using micro:bits are also described.
Art CT in Art.Students redesign a game, including both coding and aesthetic features (Bowden, 2019;2020).
Civics CT in Civics.An interdisciplinary activity (Civics and Mathematics) contextualises mathematics when the students analyse voting results with programming to learn about the Swedish electoral system (Sjöberg et al., 2019).
Craft CT in Craft.Ahmed et al. (2019) study students' perceptions of programming.
Students associate programming with Craft.This association is explained by the fact that construction is a prominent element in programming and CT.
Language CT in Language.There are several languages in the curriculum.Compulsory languages are Swedish/Swedish as a second language and English.Students associate programming with language, as there are similarities in structures and syntax.They also associate with language when using Swedish and English in programming (Ahmed et al., 2019).Heintz et al. (2015)

Discussion and conclusion
The curriculum is an expression of educational policy governance and formulates desired goals for education (Larsson & Westberg, 2019).Publications from the National Agency for Education claim that CT is important for understanding programming (Skolverket, 2018c), yet our analysis is that this is not reflected on a semantic level in the curriculum.While the content analysis only focuses on a semantic level and does not provide interpretations of underlying meanings, the results offer an understanding of the conditions for integrating CT into both the curriculum and practical implementation.
The results indicate a strong emphasis on digital competencies in the curriculum, with a particular focus on digital tools and media (Figure 2; Table 3).However, compared to the pervasiveness of digital competence in the curriculum, programming receives only moderate attention, and CT is entirely absent (Figure 2; Table 3).Therefore, we conclude that the conditions for CT to emerge as a priority subject are poor.
On the other hand, the research literature underlying the thematic analysis shows that CT is already integrated into classroom practice through programming activities (Table 5).Characteristically, the activities that include CT are programming using Scratch software.Although the reviewed literature offers various descriptions of CT, a unified definition is still difficult to achieve.Nevertheless, our analysis reveals four broad definitions: components, concepts, approaches, and practices (Appendix 3).There are also indications that CT continues to evolve.Researchers build on definitions of other researchers (Zhang et al., 2020), and strategies regarding assessment and progression being developed (Mannila et al., 2020;Zhang et al., 2020).Despite an increasing diversity of CT definitions, the legacy of Papert and Wing is clear (Figure 3).Their influences are also reflected in Themes (A), (B) and (D) in Table 4.These themes summarise the essence of CT as a universal competence useful for problem solving.Theme (C) highlights the interconnectedness between CT and programming activities.
We also see a contradiction in the results: Theme (D) presents CT as a universal competence and Theme (F) as domain dependent and linked to different school subjects (Table 4).This contradiction can actually help us to pinpoint CT.CT is a competence for everyone, not just for computer scientists as Wing claims (2006;2017).It is also domain dependent, which is in line with Denning and Tedre (2019): "We see CT as mostly domain dependent -for example, how you think about computation in biology is different from physics" (p.191).Despite the weak support from the curriculum and the diversity of CT definitions, the results of our thematic analysis show potential for using CT in specific subjects.Not only in subjects with programming objectives but in others as well.The results in Table 5 show examples of CT adapted to domain-specific use.
However, there can be problems with limiting CT to programming and problem solving.CT is changing and expanding to focus on the design of complex systems where a deep understanding of humans and technology are equally important (Denning & Tedre, 2019;KTH, 2022;Pears et al., 2021).Meeting these demands in education can be a challenge, but one subject in the Swedish curriculum seems to be suitable for the task, the Technology subject.The subject is already about the artificial world, the creation of technical artefacts and systems, ethical dilemmas in sustainable product development, and historical and future perspectives on technology (Skolverket 2018a;2018b).The subject is characterised by engineering methods as the design process with a strong focus on human-centred design (Skolverket 2018a;2018b).The design process is about finding the right problems, solving them by constructing artefacts, and thus fulfilling human needs (Hughes, 2009;Norman, 2013).It also seems to be the most appropriate school subject to take responsibility for the legacy of Papert and further develop his ideas on teaching strategies for the interplay between humans and technical artefacts, where the students' own ideas are at the centre.Based on our previous arguments, we believe that the Technology subject offers significant potential as a research arena for exploring how to integrate CT into engineering approaches, including the design process and system design.
We conclude that Sweden, and other countries with a similar integrative approach to embedding CT in the curriculum, need to place greater emphasis on the relevance of CT and programming for school subjects like Technology, where programming is explicitly mandated to be taught, and where CT skills can be expected to be developed.We argue that the Swedish curriculum must be strengthened, providing a clearer role for the subject of Technology in developing CT and programming.

5 SCOPUS
2021(All years).References ERIC-education AND computational thinking AND SwedenAustralasian Journal of Technology Education, Special Issue: Technology Education on the Edge, Vol. 9, 2023 Years 2015-2021 (All years).References (school AND computational PRE/0 thinking AND Sweden screening titles -STEP 2. Identification, reading abstracts -STEP 3. Removing duplicates -STEP 4. Read full texts -STEP 5. Inclusion, selections of texts Besides the keywords, the research should concern teaching and learning from preschool to upper secondary school.The earliest study meeting these criteria was published in 2015 and the latest in 2021, the same year the search for publications ended.

Figure 1 .
Figure 1.The seach and selection procedure for the rapid literature review.

Figure 2 .
Figure 2. References counted in the content analysis referencing each framework Australasian Journal of Technology Education, Special Issue: Technology Education on the Edge, Vol. 9, 2023 11

Table 5 shows
CT related to school subjects.In Mathematics CT relates to problem solving and to automation in Technology.Few results in the publications identified CT in relation to Civics apart from a description of a cross-disciplinary project with Mathematics.CT was also identified in subjects lacking programming learning objectives.In Physics, the students worked with simulations and in Art with visual interfaces.CT was also associated with Craft and Languages.Additional examples of classroom activities including CT were game design, animation, and visualisation.A common activity was block-based programming with Scratch software.But other resources besides Scratch was found in the analysis: programming languages e.g., Python; robots e.g., BlueBots and Lego EV3; embedded systems e.g., Arduino and Micro:bits

Table 5 . Computational Thinking Related to Swedish School Subjects Subjects Computational thinking (CT) related to school subjects. Summaries.
Mathematics CT in Mathematics is discussed in relation to problem solving.