Computational Thinking in Education: Where does it Fit? A systematic literary review

1 Introduction

• Wing describes CT as solving problems, designing systems, and understanding human behavior using computer science concepts, and says it requires more than programming.The introduction notes that academics have not agreed on a universal definition.Source: §1 · p. 4
• In Ireland, students cannot yet take Computer Science or Computing as a state-exam subject; the available curriculum offering is a coding short course.The introduction says steps toward inclusion have been taken.Source: §1 · p. 4
• The PACT program introduces Irish secondary students and teachers to Computer Science through programming and algorithms, and has reached over 60 schools and 1,000 students.PACT stands for Programming ^ Algorithms = Computational Thinking.Source: §1 · p. 4
• The review examines how other countries and institutions have incorporated CT, with the stated aim of informing educators, researchers, teachers, and industry members.The introduction links this aim to Ireland’s plans for a Computer Science curriculum in the final two years of secondary education.Source: §1 · p. 4

1.1 Note on education terms

• Education-level labels are approximate: secondary schooling commonly spans ages 11/12–17/18, while college and university refer to tertiary education.The passages note that age ranges and institutional terms vary between countries.Source: §1 · p. 5

2 Research questions

• The review asks what methods, tests, and tools exist to assess students’ CT ability or improvement.Source: §2 · p. 5
• It asks why educational institutions should incorporate CT into curricula and what benefits CT offers.Source: §2 · p. 5
• It asks how CT has been incorporated into existing subjects and courses.Source: §2 · p. 5
• It asks what tools have been developed or used to teach CT.Source: §2 · p. 5
• It asks how CT has been taught in educational institutions.Source: §2 · p. 5

3.1 Introduction

• The review uses Kitchenham’s systematic-review method, which the authors chose because it is documented and derives from established medical review processes.The method covers identifying the need for a review, developing a protocol, and reporting findings.Source: §3.1 · p. 5

3.2 Search terms

• The search uses “Computational Thinking” as its primary term, combined with education-level, institution, and assessment terms.Source: §3.2 · p. 5

3.3 Resources searched

• Searches ran in October–November 2016 across ACM, IEEE Xplore, ERIC, and Google Scholar, with papers retained after title-based screening.ACM returned 363 papers and retained 160; IEEE returned 107 and retained 63; ERIC returned 9 and retained 7; Google Scholar yielded 882 papers and retained 136.Source: §3.3 · p. 6

3.4 Document selection

• Studies had to address a research question and concern CT teaching or relevant problem-solving and analytical skills in educational courses.The review excluded books and grey literature, non-peer-reviewed or irrelevant work, and non-English texts; 210 papers remained after abstract screening.Source: §3.4 · p. 6

3.5 Quality Assessment

• The authors assessed eligible papers using questions adapted from Kitchenham’s framework about evaluation, data collection, reporting, and research contribution.Thirty-five papers were rejected during quality assessment.Source: §3.5 · pp. 6–7

Teacher preparation

• Reports identify teacher preparation as a challenge, including shortages of trained CS teachers and limited teacher education, while training programs help teachers incorporate materials into classes [16].The reported evidence was not sufficient to establish the same impact on students [16].Source: §4 · pp. 8, 10

Programming and teaching approaches

• Reviews describe programming as a route into CT and report low-floor, high-ceiling languages, visual outputs, scaffolding, reflection, and reinforcement as teaching approaches [9].Reported uses include learning languages and mathematics and teaching concepts such as variables and loops [9].Source: §4 · p. 8

Tools and instructional practices

• A review of K–12 computing education identified 33 tools and found game design, tangible construction kits, and robotics popular, while noting that differing class features make findings difficult to compare.Source: §4 · p. 9

Curriculum integration

• Prior work presents CT through mathematics and other subjects, standalone courses, and integrated curricula, while noting that further curriculum integration is needed.One review connects CT with mathematical representations, algorithms, pattern regularities, and models.Source: §4 · pp. 9–10

Related reviews

• Prior reviews find no widely accepted CT definition and describe the research as being at an early stage.A review of 125 papers found 43 concerned CT integration into education, courses, or activities.Source: §4 · p. 9

Higher education

• Higher-education examples use programming, MATLAB, robotics, and realistic team projects to teach CT, with recommendations to integrate tools across courses and departments.The review of college teaching also characterizes CT as valuable for CS and non-CS students.Source: §4 · pp. 10–11

• Seiter and Foreman’s PECT model scores Scratch code blocks by complexity and evaluates design-pattern use; preliminary results showed CT skills increasing with creators’ age.PECT separates CT concepts, Design Pattern Variables, and Evidence Variables.Source: §5 · p. 11
• Dr. Scratch scores correlated moderately to strongly with established software-complexity metrics, but its CT scores could differ across Scratch project types and from qualitative assessments.Moreno-leon et al. compared the tool with McCabe’s Cyclomatic Complexity and Halstead’s metrics; reports differences across project types, and Hoover et al. found quantitative and qualitative results differed.Source: §5 · p. 11
• Sherman and Martin’s rubric assesses 14 properties in App Inventor projects, combining general computational thinking with mobile computational thinking concepts.Two authors rated 45 apps from 18 students on a 2-to-4-point scale.Source: §5 · pp. 11–12
• Basawapatna et al. developed a quiz in which students identify computational thinking patterns in videos and propose patterns for a predator-prey simulation.The quiz addressed whether these patterns are accessible and useful to teachers and students in game design.Source: §5 · p. 12
• Koh et al. developed REACT to show teachers students’ implemented, unused, and incorrectly implemented patterns during coding; four teachers responded overwhelmingly positively after testing 134 projects.The authors describe the system as a first step and note that further work and study are needed.Source: §5 · p. 12
• Basu et al. evaluate student models against expert models and use model accuracy, effectiveness, and consistency measures to assess computational thinking in simulation activities.Effectiveness—the proportion of edits that improve model accuracy—was a strong predictor of final model accuracy.Source: §5 · p. 13
• CT pass rates were 55.4% vs. 85.5% course pass rates in Gouws et al.’s assessment, and results showed a varied relationship between CT and class tests.The test used questions from Computer Olympiad ‘Talent Search’ papers and was administered before and after a semester of computer science study.Source: §5 · pp. 13–14
• Werner et al.’s Fairy Assessment used three independent Alice tasks to assess algorithmic thinking, abstraction, and modelling; only 30 of 311 students did not attempt to modify the program.Task 2 was the most complex and had the lowest mean scores.Source: §5 · p. 14

5.2 Conclusion

• Lishinski et al. found problem-solving scores predicted constructed-response programming outcomes, but not multiple-choice exam results.They also report that problem-solving skills uniquely predict higher-level skills beyond lower-level skills.Source: §5.2 · p. 14
• Haddad and Kalaani found CT performance correlated with GPA, while Oliveira et al. found significant correlations between computing-test and academic scores across primary-school classes.Source: §5.2 · p. 15
• Van Dyne and Braun report analytical-skill scores rising from 60.3% before to 78.4% after a CT-focused CS0 course for majors and non-majors.The authors report an increase of over 18%.Source: §5.2 · p. 15
• Davies found students taught structured pseudocode for most of a course performed as well on a final exam as those taught C++ for 13 weeks.The formal language was introduced only in the final three weeks; the authors also report that women might benefit in attitudes and confidence.Source: §5.2 · p. 15

6.2 Conclusion

• Teaching CT or integrating its concepts could improve analytical skills, clarify programming as problem-solving, support women’s programming confidence, and correlate with academic success.Source: §6.2 · p. 15
• CT research remains early-stage, and long-term effects and additional benefits still need investigation.The review notes a need for more detailed assessment work, as many testing examples are at early development stages.Source: §5.2 · p. 14, §6.2 · p. 16

7.2 Sciences

• When standalone CS is impractical, CT can be integrated across subjects; Swaid proposes a hands-on course that first teaches CT concepts and applications, then Java programming.Source: §7 · p. 16, §7.2 · p. 17
• Hambrusch et al. report increased interest in another CS course and positive feedback on a problem-driven science course; Sabitzer and Paster report positive evaluations of COOL Informatics pilots.The COOL Informatics approach used informatics concepts and hands-on activities to support learning across subjects; a teacher workshop also improved participants’ CT understanding and use of computational tools.Source: §7.2 · pp. 16–17

• Qin mapped CT concepts to bioinformatics topics and used hands-on learning; post-course surveys found students felt their computer knowledge and quantitative skills had improved.Students worked in pairs, with quizzes, changing pairs, and discussions used to support individual learning.Source: §7.3 · p. 18
• Rubinstein and Chor mapped biological domains to computational topics and concepts, including networks to graphs and Dijkstra’s algorithm to greedy algorithms, abstraction, and reduction.Source: §7.3 · p. 18
• Rubinstein and Chor [45] report that students solving real-life Python problems focused on computational concepts rather than syntax, but their evaluation was not in-depth.Students’ views of computer science shifted from the physical machine toward problem-solving and other CT concepts.Source: §7.3 · p. 18

7.5 Maths

• A high-school math workshop used abstraction, generalisation, justification, and Python mini-programs; teachers saw programming as a tool for teaching mathematical reasoning.Teachers were also eager to learn more about programming and its classroom integration.Source: §7.5 · p. 18
• Polish textbook topics absent from mathematics texts were linked to mathematical ideas, including number representation to polynomials and heuristic thinking to shortest paths.Other examples include triangle validity, approximation and rounding errors, and recursion.Source: §7.5 · pp. 18–19
• The MPCT introductory programming course used calculus-based modules to program algebra and geometry concepts, aiming to build intuition, appreciation, and confidence in pre-calculus.It was designed for students entering their freshman year.Source: §7.5 · p. 19

7.6.1 Journalism

• Interactive Journalism students used Scratch, text, and video to create news stories; authors reported increased confidence but said whether the projects demonstrated CT was difficult to determine.Students had the most fun during the Scratch programming sections, and the authors saw similarities between good writing and animation design.Source: §7.6.1 · p. 19
• Journalism students rated their CT skills higher after a five-semester collaboration with CS students, with statistically significant post-course results across all measures.The students jointly developed a website for Habitat for Humanity, dividing technical work and journalism research and writing.Source: §7.6.1 · p. 19

7.6.2 Writing

• English faculty used concept maps of writing structures to make implicit knowledge of writing concepts more explicit; classroom transfer would require teaching students to use the maps.The project was an initial collaboration between English and Computing departments.Source: §7.6.2 · p. 19

7.6.3 Literature

• DISSECT students performed better than control classes on CT assessments and a writing test, and their classes demonstrated CT skills such as abstraction.The review reports the integration as successful.Source: §7.6.3 · p. 20

7.7 Dance

• In dance programming, students used modularisation, testing, and debugging to build and refine choreography; interviews suggested changed computing perspectives and enjoyment.The authors describe the work as an initial stage and say future work is needed.Source: §7.7 · p. 20

7.8.1 Framework - general

• DePaul revised 19 general education courses to make CT more explicit and described teaching, learning goals, and assessment in three detailed modules.Examples included geographic information systems, game design, and 3-D modelling.Source: §7.8.1 · p. 20

7.8.2 Interdisciplinary

• A three-department gerontechnology course focused on CT across design, Computer Science, and gerontology; students reported improved CT competency and interdisciplinary teamwork.Non-CS students performed comparably to CS students in CS modules, while reported competency to engage in CT declined.Source: §7.8.2 · p. 20
• Teachers integrated computational activities into existing art, biology, health, geography, and civics courses; preliminary reports described increased teacher interest and K-12 student engagement.Teachers planned to continue incorporating the computational content.Source: §7.8.2 · p. 21
• A proposed age-group framework matches CT skills and teaching examples to stages from K-2 through grades 9-12.Examples range from pattern recognition and LOGO/Scratch activities to programming languages and troubleshooting.Source: §7.8.2 · p. 21

7.9 Conclusion

• The review concludes that CT can be taught in varied subjects without requiring a new course or exclusive placement in Computer Science.The authors connect this flexibility to already crowded curricula and CT's potential usefulness across students' academic and work lives.Source: §7.9 · p. 21

8.3 App Inventor

• App Inventor supported mobile-app exercises that introduced CT practices, IT-course modules that students rated highly, and discussion-centered app building that showed preliminary CT growth [81].The MobileDev summer school used increasingly difficult exercises and an Agile final project; its authors informally observed student enjoyment. In the college IT course, app development scored almost 90% on survey responses about CT use; the seven-student discourse-focused pilot observed increased CS vocabulary and CT elements.Source: §8.3 · p. 24

8.4.1 RaBit EscApe

• RaBit EscApe uses magnetized wooden pieces to teach path-making, and an informal trial suggested scaffolding difficulty and a cheat sheet could improve this tangible CT activity.The game was designed for ages 6–10; two groups of three children aged 8–10 played it.Source: §8.4.1 · pp. 24–25

8.4.2 Board Game - Pandemic

• In Pandemic sessions, distributed computation was the most frequently observed CT concept, leading the authors to propose strategic board games as a foundation for intentional CT learning.The study coded statements from three groups of first-year undergraduates for conditional logic, algorithm building, debugging, simulation and distributed computation.Source: §8.4.2 · p. 25

8.4.3 Program your robot

• “Program your robot” maps CT skills to an algorithm-building game, but its informal CS-student feedback is not empirical evidence that the game teaches CT.Students enjoyed the game and suggested it could enhance problem-solving in introductory programming courses.Source: §8.4.3 · p. 25

8.4.4 Agentsheets/AgentCubes

• The Scalable Game Design work reported transferable problem-solving and sustained school participation, while its CT-tool criteria include a low learning threshold and a high sophistication ceiling.Over 80% of schools advanced beyond the first project, and 43% moved from game design to science simulations.Source: §8.4.4 · p. 26

8.5 Lego Mindstorms

• Lego Mindstorms supported robotics-based CT instruction: studies reported improved analytical or CT skills, and TangibleK children met target achievement levels.Atmatzidou and Demetraidis found students reached the same CT skill level regardless of age, while girls in the junior-high group took longer. TangibleK students found the more complex second half harder, but scored higher on choosing and sequencing instructions in the final project.Source: §8.5 · p. 27

8.6 CS Unplugged

• Bell et al. described CS Unplugged adaptations including activity videos, binary-coded bracelets and activities themed around subjects such as WWII.They also discussed activity-design principles and future plans.Source: §8.6 · p. 28
• In “Marching Orders,” fourth graders recognized the need for specific instructions but struggled to produce them, and saw that small errors could change outcomes.The activity has one student direct another to draw a shape.Source: §8.6 · p. 28
• CS Unplugged lessons showed uneven proficiency across topics: binary numbers exceeded 80% on all six questions, while searching reached 53%, and classroom use needed enhancement.Suggestions included a naive first attempt, individual practice, consistent worksheet vocabulary and engaging real-world concepts.Source: §8.6 · p. 28
• A course training undergraduates and trainee STEM educators used Scratch, CS Unplugged and HTML to teach CS and CT, with participants noting that unplugged activities require practice and scaffolding.Undergraduates taught school classes involving concepts such as variables, loops and conditionals.Source: §8.6 · p. 28
• After CS Unplugged activities, seventh graders’ CS intentions and attitudes did not improve, their view of CS as fun or interesting declined, and their understanding of CS improved less than hoped.Students already viewed CS careers for women positively; the authors suggested activity order and unclear links to high-school CS or careers might explain the decline in interest.Source: §8.6 · pp. 28–29

8.7 Other tools

• The review found over 50 CT teaching tools, including Arduino and CTSiM.The list also names tools such as the Drawing Machine Model, Greenfoot, MATLAB and The Incredible Machine.Source: §8.7 · pp. 29–30

8.8 Conclusion

• CT teaching tools range from music tools to programming languages and games, offering options for students of different ages, genders and abilities across varied classroom settings.Some tools remain at early development stages, but the review describes the breadth of options as encouraging for educators.Source: §8.8 · p. 30

9.1 Introduction

• Computer Science or Computing courses, classes, and workshops are described as the most common way CT is introduced.The section introduces a broad range of such offerings and how schools and higher-education institutions developed standalone CT modules and courses.Source: §9 · p. 30
• The section aims to offer ideas, encouragement, and resources to educators with existing CS classes or the opportunity to teach CT independently.Source: §9 · p. 30
• The review situates these courses alongside previously discussed tools and ways to incorporate CT into existing subjects.Source: §9 · p. 30

9.2 Teacher workshops

• Over 90% of teachers planned to use ACTIVATE workshop materials in new or existing courses.The week-long workshops combined programming and computational concepts with an application teachers designed around topics they already taught.Source: §9.2 · p. 31
• 75% to 89.65% of teachers correctly identified CT after Morreale et al.’s first workshop, while correct understanding of its importance rose by over 20%.Four of eight workshop sessions were identified as immediately useful; the most used concerned CS careers, with others on algorithmic thinking and CS Unplugged.Source: §9.2 · p. 32
• Workshops using Backwards Design helped participants clarify CT concepts, but participants mainly focused on one of the eight CS Principles framework areas.Participants designed and presented lesson plans using at least one learning objective.Source: §9.2 · p. 32
• Workshops were associated with improved teacher understanding of CT and CS, though follow-up also exposed practical barriers to classroom use.Teachers reported difficulties including limited IT support and time to refine lesson plans or develop subject competence.Source: §9.2 · p. 32

9.2.1 Conclusion

• Practical, often day-long workshops can improve teachers’ enthusiasm, knowledge, and ability to teach CT, while misconceptions and teacher willingness remain important implementation factors.The review describes wide-reaching initiatives such as CS4HS as support for teachers who otherwise lack relevant skills and knowledge.Source: §9.2.1 · p. 33

9.3 Early Education

• Lu and Fletcher argue that programming should not be the first thing taught; they propose introducing vocabulary and symbols for computation and abstraction through a computational thinking language.They offer examples including multiplication, charting information, square roots, and interdisciplinary projects.Source: §9.3 · pp. 33–34
• Video interviews of second-grade children explaining ScratchJr projects captured evidence of concepts including reuse, sequencing, and space cost.The authors recommend practice, camera-work, and presentation demonstrations before conducting the paired interviews.Source: §9.3 · p. 34

9.4 Primary School

• CargoBot and Pyonkee appeared useful for different CT dimensions: CargoBot supported more technical concepts and more testing and debugging, while Pyonkee appeared to prompt more collaboration.With both apps, little trial and error was observed, suggesting problem solving and debugging were deliberate and reflective.Source: §9.4 · p. 34
• After a year of Logo experience, fifth- and sixth-grade treatment groups differed significantly from controls on both problem-solving measures and one of three spatial tests, but no pre-tests were given.Without pre-tests, the study cannot establish which group learned more.Source: §9.4 · p. 34
• A primary-school CS/CT course offered opportunities for cross-curricular teaching in Maths, PE, Creative Writing, and Art, with teachers also noting teamwork, cooperation, and communication.Teachers reported that they could deliver the material in an engaging manner, often better than expected.Source: §9.4 · p. 34
• Macedonia introduced a computing subject for eight-year-olds, including CT games and ScratchJr programming; trained teachers became more confident and understood the topic’s need.The subject was scheduled for two classes per week and included seven units.Source: §9.4 · p. 35
• Primary-school lessons also used a self-contradictory dollhouse to teach affordance and negative transfer, and recipe-symbol activities to teach recognition in design.The sessions could add iterative design through second and third attempts.Source: §9.4 · p. 35

9.4.1 Conclusion

• The review concludes that visual programming such as Scratch or ScratchJr and hands-on activities such as CS Unplugged offer ways to teach CT to primary pupils.It also stresses showing teachers how CT fits other subjects or may already be present in their teaching.Source: §9.4.1 · p. 36

9.5 Secondary School

• A high-school curriculum for Israeli middle schools aimed to teach logical and algorithmic thinking, not to make students programmers, through four modules over three years.The course totaled 180 hours, with two hours per week.Source: §9.5 · p. 36
• Brazilian implementation discussions identify infrastructure, curriculum organization, instructor qualifications, dissemination strategies, and government policy as challenges to including CT in schools.The authors also note possible reluctance where people believe they already possess the relevant skills.Source: §9.5 · p. 37
• Secondary-school CT activities included binary search, quick sort, and Scratch iteration; students demonstrated decomposition, but found one activity boring despite finding it helpful.The first two activities were described as interesting and helpful to learning.Source: §9.5 · p. 37
• DISSECT modules generally increased students’ interest, motivation, and CS knowledge, while separate module studies reported statistical CT-knowledge gains; the authors note that further study of concept mastery is needed.Qualitative findings described students understanding concepts and enjoying the modules.Source: §9.5 · p. 38
• FACT assessment and interview data indicated improved understanding of algorithmic flow, and its online version worked as well as or better than its face-to-face version [74].After seven weeks using Scratch, students could interpret text-based programs; the authors suggest learned algorithmic concepts may transfer to other tools.Source: §9.5 · pp. 38–39
• The review describes diverse approaches, including curricula, optional courses, teacher development, cross-curricular projects, and braided teaching that interweaves topics rather than prioritizing one.Examples include introducing CT through chemistry, physics, and games, and teaching programming alongside other topics.Source: §9.5 · p. 39, §9.5.1 · p. 40

9.5.1 Conclusion

• The review identifies shortages of skilled or trained teachers and equipment, especially in poorer countries, alongside a wealth of available teaching resources and tools.It notes that different tools and concepts suit different age groups, such as Scratch for younger students and higher-level programming later.Source: §9.5.1 · p. 40
• The review concludes that efforts are making CS and CT more accessible, widespread, and teachable in secondary schools.Across the reviewed work, concepts and topics are presented as ways to develop CT skills and students’ enthusiasm for and understanding of CS.Source: §9.5.1 · p. 40

9.6 College/University

• Astrachan uses active lectures and discussion rather than hands-on work, while Czerkawski maps CT concepts to online activities such as surveys, data analysis, and instructional design.The online activities introduce computational vocabulary and mental tasks without programming; social collaboration tools may ease a later transition to programming.Source: §9.6 · pp. 40–41
• Dierbach et al. developed four discipline-specific undergraduate courses around shared CT objectives; their initial assessment found faculty interest, student perceptions of relevance, and positive administrator feedback.The objectives include applying computational models and algorithms and evaluating their use in disciplinary problems.Source: §9.6 · pp. 41–42
• Ater-Kranov et al. report that their multidisciplinary course engaged and motivated students, fostered cross-disciplinary learning, and deepened appreciation of computing in students’ disciplines.The project used tools including NetLogo, Python, and Blockly, with the collaborative Rhinestone e-textbook platform.Source: §9.6 · pp. 41–42
• Computational Creative Exercises improved CT and CS learning; participants also reported higher self-efficacy and study time and lower lack of regulation than the control class.The quasi-experimental study involved 90 students and a control class of 65.Source: §9.6 · pp. 42–43
• Miller et al.’s [195 and 196] creative exercises appeared to improve achievement, retention, and grades, although students did not always recognize the benefits of creative thinking for CS.The authors call for clearer explanation of how creative thinking relates to CS topics.Source: §9.6 · p. 43
• Cortina designed an introductory algorithms-focused CS course for non-majors without programming; over half would consider another CS course and 85% would recommend it.Students used Raptor, a flowchart-building tool, to see algorithms work.Source: §9.6 · p. 43
• Across college contexts, CT activities include data-driven MATLAB work, networking experiments linked to CT concepts, and courses assessing students’ understanding of algorithms, abstraction, and efficiency.The biology case study involved five undergraduates; Qualls et al. found students could illustrate abstraction despite difficulty defining it, and could relate efficiency to other fields.Source: §9.6 · pp. 43–44

9.6.1 Conclusion

• The review concludes that practical, discussion-led college courses may benefit CS and non-CS students, while noting that success requires support from administrators and teaching staff.It suggests non-compulsory CT courses may be particularly effective and that they might ease CS students’ transition to traditional programming.Source: §9.6.1 · p. 44

10 Conclusion

• The review aims to give educators an overview of current CT teaching research and classroom examples across educational contexts.It frames CT as a skill to teach students of all ages.Source: §10 · p. 44
• CT can be taught through dedicated modules, existing subjects, or one-off events, using varied tools and approaches.The paper describes these approaches as potentially engaging and applicable across curriculum, daily life, and work.Source: §10 · p. 44
• The conclusion notes that CT assessment approaches exist, but more work and more detailed lesson plans and curriculum structure could benefit teachers.Source: §10 · p. 44