Open research questions in Teaching and Learning Programming

What the literature leaves open

• Prior research suggests that POI supports students' abstraction, analogical reasoning, and decomposition skills, though empirical evidence remains limited.

  Abstraction First: Re-examining the Role of Algorithmic Patterns in CS Students' Abstraction Development · 2026 · DOI

• Open question: Describe the difference between Continuous Integration and Continuous Deploy- ment. a) Strong, as it is supported by research b) Weak if there is no empirical evidence c) Neutral if there are mixed results 286 ISSN 1392-5016 eISSN 1648-665X Acta Paedagogica Vilnensia 56, 2026 5. You have received two conflicting reports on CI/CD performance.

  Issues with Using Artificial Intelligence in an Educational Environment: Implications for the Training of DevOps Professionals · 2026 · DOI

• However, research remains fragmented, as most studies examine a single pedagogical modality in isolation, leaving limited evidence on how young children progress across different types of CT experiences.

  Computational Thinking Pathways in Preschool: A Multimodal Approach · 2026 · DOI

• @ese findings highlight the potential of the rubric to support less experienced teachers and suggest that future research should explore strategies to further facilitate its effective use. Existing approaches vary widely and are often designed for standardized, task-based contexts rather than authentic, open-ended projects. Currently, there is no standardized, empirically validated framework for the evaluation of block-based pro- gramming projects.

  Rubric for the qualitative assessment of student-designed Snap! projects · 2026 · DOI

• This is why universal statements about “ethical AI” are insufficient. Longitudinal and cross-cultural studies remain scarce, and few studies examine how AI- related professional learning develops into sustained classroom practice over time (Holmes & Tuomi, 2022; Niloy et al.

  Co-Teaching with AI: Shift Pedagogy, Opportunities, and Challenges in Mathematics Education · 2026 · DOI

• Furthermore, this study examined five specific CT components; future research might explore additional components such as parallelization or automation.

  The Effect of Problem-Solving-Based Programming Training on Computational Thinking Skills in the Context of Programming and Reasoning · 2026 · DOI

• The following recommendations were made based on the findings of the study: 1. Since algorithmic thinking has been found to have a strong predictive power, educators should integrate more algorithm-focused exercises, such as structured coding challenges and debugging tasks, into the curriculum to enhance logical problem-solving skills. -105- Journal of Technology and Science Education – https://doi.org/10.3926/jotse.3402 2. Decomposition was found to negatively correlate with achievement. Therefore, instructional methods should emphasise effective problem breakdown and reconstruction through guided exercises, scaffolded learning, and project-based tasks. 3. Practical and collaborative learning should be promoted to strengthen abstraction, evaluation, and generalisation. Hands-on projects, peer programming, and real-world applications should be incorporated to reinforce computational thinking in robotics programming.

  Exploring the relationship between computational thinking dimensions and achievement in robotics programming among computer education students · 2026 · DOI

• This study offers significant insights into the predictive influence of computational thinking on students’ performance in a project-based robotics programming course; however, a number of limitations were acknowledged: 1. The sample size (N=105) is relatively small and comes from just one region (Southeast Nigeria), which may make it hard to generalise the results to other regions of Nigeria or other educational systems. 2. The study utilised a correlational design, which facilitates identification of relationships among variables but does not confirm causation; hence, caution is warranted in interpreting the directionality of the observed interactions. 3. While validated instruments were utilised for measuring computational thinking, some of the assessments depended on how students perceived themselves, which could be affected by biases like wanting to look good or not knowing much about themselves. 4. The findings underscore the negative impact of decomposition difficulties on students’ achievement. Subsequent research should investigate whether structured training in decomposition strategies may mitigate this issue and improve the ability of learners to effectively integrate decomposed components into functional robotic systems.

  Exploring the relationship between computational thinking dimensions and achievement in robotics programming among computer education students · 2026 · DOI

• for Researchers Researchers should use a mixed-methods approach to gain a holistic under- standing of STEM integration, capturing student perceptions and experiences to provide vital context to quantitative findings.

  Building Beyond the Robot: Disciplinary Integration in Challenge-Based Learning · 2026 · DOI

• Use experimental designs, diverse samples, and quantitative studies to assess CBL’s effects on STEM integration. Longitudinal studies are also recommended to track students’ skill development.

  Building Beyond the Robot: Disciplinary Integration in Challenge-Based Learning · 2026 · DOI

• Future research could address these limitations by incorporating multiple program- ming languages, expanding the range of covered topics, and exploring alternative student-AI inter- action models. Additionally, the study did not examine the effects of allow- ing students to actively generate erroneous code using the chatbot, an area that warrants further investigation. Furthermore, the possible implications of inte- grating AI chatbots into programming education are not limited to their use in individual learning settings.

  AI-Generated Errors as a Learning Tool: Improving Programming Education Through Error Correction · 2026 · DOI

• Future research could address these limitations by incorporating multiple program- ming languages, expanding the range of topics covered, and exploring more interactive forms of student-AI engagement. First, the study was limited to a single programming lan- guage, which may restrict the generalizability of the findings to other languages with different syn- tax and structures.

  AI-Generated Errors as a Learning Tool: Improving Programming Education Through Error Correction · 2026 · DOI

• Abstract As artificial intelligence coding assistants (AICAs) are increasingly adopted in K–12 coding education, the role of students' computational thinking (CT) levels in the learning process remains under-explored.

  Exploring the effect of computational thinking levels on students' learning performance, cognition, and behavior when using AI coding assistants · 2026 · DOI

• LIMITATIONS This study provides a comprehensive overview of research on visual and block-based programming tools in K–12 AI education; however, there are several limitations to consider.

  Pedagogical Impacts of Block-Based Artificial Intelligence Applications: A Systematic Review · 2026 · DOI

• The approach in this study shows promise but needs to be validated, and we suggest a deeper investigation of a subset of the courses investigated, looking closer at the course material for those courses to form a clearer connection be- tween the data collected through the survey, the activities undertaken by students, and how those activities relate to subtle concepts. But this difference needs to be studied further and validated. While some findings in this paper need to be validated, we suggest the study process de- scribed in this paper as an approach to better understand the development of these types of concepts.

  Three Programs, Three Years, and Four Concepts: Teachers’ Views on Indirection, References, Scope, and Parameter Passing in CSED · 2026 · DOI

• By quantita- tively modeling the structural association between CT and SPS, this study addresses an important gap in the literature, which has predominantly focused on conceptual discus- sions or intervention-based research rather than structural and model-driven analyses (Paraskevopoulou-Kollia et al.

  An Investigation of the Relationship between Computational Thinking and Scientific Process Skills in Pre-schoolers: A Correlational Study · 2026 · DOI

• Objectives The present study addresses this gap in the literature by exploring how different human-LLM interaction modes (standard-prompting mode, user-interface mode, context-based mode, and agent-facilitator mode) can be applied in CT studies and identifying the challenges that may arise in such applications.

  A systematic review of human-LLM interactions in computational thinking empirical studies · 2026 · DOI

• Originality/value Given the freedom that higher education instructors and professors typically have to design their courses at both the undergraduate and graduate levels, relatively little is known about how students experience alternative types of pedagogical approaches and assessment methods.

  Student perspectives on grading in higher education STEM courses · 2026 · DOI

• This study has three methodological limitations. The first concerns the handling of the ideas in the discourse data during KBDeX analysis. Although context-sensitive judgment was applied when identifying representative ideas in the final presentations, such seman- tic judgment was not extended across the entire discourse. Japanese morphological analyzers segment compound nouns into constituent morphemes, which can result in Naganuma et al. International Journal of Educational Technology in Higher Education (2026) 23:24 Page 27 of 31 semantically unrelated morphemes being treated as identical lexical units. For example, the compound shūkyōshoku (宗教色, “religious nuance”) is decomposed into shūkyō (“religion”) and shoku (“color”), even though shoku does not retain its literal meaning in this context. As a result, minor semantic noise may be introduced into degree-cen- trality calculations. While similar issues could theoretically occur in English, they are more pronounced in Japanese due to morphological segmentation. Nevertheless, this approach was deemed appropriate because the primary aim of the KBDeX analysis was to capture structural patterns of idea improvement with objectivity and reproducibility rather than to interpret the semantic intent of each occurrence. The second limitation concerns the inferential identification of idea sources in group discourse. Source attribution was based on temporal relationships between students’ utterances and AI usage records, allowing only plausible, not definitive, inferences. For example, when an idea articulated by Speaker A appeared earlier in Speaker B’s AI usage history, we treated Speaker B’s AI as a possible idea source, even if the idea did not appear in Speaker A’s own AI records. This inference was informed by classroom observations indicating frequent screen sharing during groupwork, which made cross- exposure to AI outputs plausible. However, because the dataset does not provide direct evidence that Speaker A actually viewed Speaker B’s AI output prior to the utterance, source attribution remains indirect and inferential. The third limitation relates to operationalizing sustained idea improvement, which comprises depth- and breadth-oriented dimensions (Hong & Sullivan, 2009; Hong et al., 2025). While KBDeX allowed us to trace structural trajectories of idea improvement, it could not distinguish between depth-oriented and breadth-oriented improvements, which is challenging even for human analysts. For instance, an idea “augmented reality” can be interpreted as an elaboration of technology use (i.e., depth) or as an attempt to broaden space design (i.e., breadth). Given the inherent ambiguity of such distinctions, our analysis could not fully capture the interplay between these two dimensions.

  Beyond expert knowledge toward idea innovation: potential and challenges of a generative AI-supported jigsaw method · 2026 · DOI

• Among the five key observations, Observations 2 and 4 in particular provides a critical baseline for a subsequent cycle of DBR, strongly encouraging the investigation of additional mediating processes and the corresponding reconsideration of the embodied design and higher-level conjectures. We addressed the two RQs as follows. Regarding RQ1, our observations indicate that the proposed Knowledge Expansive Jigsaw Method promoted ideation practices by enabling students to move beyond “pooling of ignorance” and engage in idea expansion through the interplay of human-generated and AI-inspired ideas in discourse. Regarding RQ2, the results indicate that the refined Knowledge Expansive Jigsaw Method did not lead to increased idea innovation. One possible explanation is that the refinement increased students’ reliance on AI-generated ideas, thereby making the relatively average and thematically convergent outputs of AI more dominant in the final products, which may have constrained the emergence of highly innovative ideas. Therefore, future work would benefit from both design refinements and improvements in the analytical approach. From a design perspective, this study has made an original contribution by demonstrating both the potential and the limitations of incorporating GenAI into collaborative knowledge-creation activities. While the GenAI-supported jigsaw method can support Naganuma et al. International Journal of Educational Technology in Higher Education (2026) 23:24 Page 28 of 31 idea expansion beyond expert knowledge (Claim 1), strengthening the mediating processes is required to achieve high levels of idea innovation (Claim 2). Such mediating processes might include promisingness judgment and regulation of idea diversity. To support learners’ capacity for promisingness judgment, one of the simplest approaches would be to allocate more time to the expert activity. A recent study conducted by Naganuma et al. (2026) demonstrated that interdisciplinary doctoral student teams could develop innovative research proposals by making promisingness judgment about AI-generated ideas. This high performance can be partly attributed to their high expertise. Therefore, if novice students such as the freshmen in this study were given more time to deepen their disciplinary understanding during the expert activity phase, they might be able to engage in effective promisingness judgments regarding expanded ideas in the subsequent jigsaw and expansion activities.

  Beyond expert knowledge toward idea innovation: potential and challenges of a generative AI-supported jigsaw method · 2026 · DOI

• Several limitations should be considered. First, the sample size was modest and drawn primarily from rural, high- poverty school contexts, which may limit the generalizability of the results. Participants were also self-selected into the professional development program, which may introduce bias toward teachers who were more motivated or open to innovation. Second, data sources included self-reported perceptions, discussion board posts, and selected artifacts, which may not fully capture enacted classroom practice. Direct classroom observations were not included. As a result, conclusions about instructional implementation should be interpreted with caution. Third, the quantitative component of the study relied on descriptive statistics and effect size estimates because matched-pair data were not available. This limits the ability to make strong inferential claims about the significance of observed differences. Finally, the study did not include direct measures of student learning outcomes. The extent to which these instructional changes influenced students’ achievement in mathematics, science, or CT remains unclear. Despite these limitations, triangulation across multiple data sources provides a comprehensive understanding of CT implementation.

  Exploring Fifth-Grade Teachers’ Integration of Computational Thinking, Instruction Strategies, and Efficacy Following Professional Development · 2026 · DOI

• Moreover, to lessen cognitive barriers, the proponent suggested the implementation of framed tasks that gradually increase in difficulty. From the results and conclusions obtained through the conduct of the study, the proponent made the following recommendations in order to enhance the programming performance and problem-solving abilities of the BSCS students. It was suggested that the programming teachers should go beyond just teaching about programming syntax and adopt more holistic strategies for teaching problem solving and algorithms. This would help learners cope with the high cognitive demands of programming by helping them handle their cognitive loads better.

  Analysing Common Barriers in Programming: Translating Real-World Problems into Functional Code · 2026 · DOI

• Future research could examine how different pedagogical approaches, such as project-based learning, inquiry-based learning, or interdisciplinary STEAM initiatives, affect stu- dents’ understanding of artificial intelligence concepts. Moreover, additional research is needed to analyze the ethical and critical dimensions of AI edu- cation, particularly regarding algorithmic bias, data awareness, and responsible technology use in school contexts.

  Teaching artificial intelligence through machine learning · 2026 · DOI

• Study Limitations While the findings of this study provide valuable insights into learning through block-based programming in the classroom, it is important to be cautious when making generalizations about the application of these findings beyond this study due to a number of limitations. The lack of comparison groups limits us from making strong conclusions regarding the effectiveness of block- based programming. It is evident from this study that students achieved significant learning via this means; however, we cannot conclude with certainty that the primary reason for this learning was due to being taught via block-based programming and not through other means. The outcomes reported were only observed over two school years. It is necessary to continue research over time to determine if this type of teaching will have a long-term impact on student success in computer science and on students’ future career choices, and whether it engenders continued interest in computing. Longitudinal studies would provide further data to indicate the extent to which students can successfully transfer knowledge gained from block-based programming into the workplace. The research team also identified limitations in generalizing their results because of how the students were grouped. All of the schools involved in the study chose to implement blockbased programming. It is possible that these school systems had greater access to technology or were already ahead of the curve in terms of using new technology in education. The findings may differ if schools had less access to technology or if they were located in an area where the community did not support technology-based education. Additionally, all the schools involved were located in the same geographic region where the cultural influences may have affected the way that students responded to programming education. In addition to that, the quality of teacher preparation was a variable even though there were efforts made to provide skills improvement opportunities to the teachers within each district. Some teachers were very Page 13 www.rsisinternational.org INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING, MANAGEMENT & APPLIED SCIENCE (IJLTEMAS) ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue V, May 2026 encouraging and involved with their students, whereas others struggled with some of the teaching formats that they were expected to use. This kind of difference is often the case in most educational settings, and can create some disparity in what the students gained from their educational experience. The official measures of student learning have not captured all of the important areas related to learning. Learning computationally involves much more than the indicators of traditional assessment. Learning the concepts of computer programming cannot be fully understood by simply managing a project. Ideally, future studies should develop better and broader assessments for computing-related skills.

  Block-Based Programming for Education: A Comprehensive Analysis of Visual Programming Environments in K-12 Learning · 2026 · DOI

• Several research paths that look promising come from this work. Studying different block-based platforms through comparative studies, could identify optimal design for various educational backgrounds. Although, early in educational programming, Scratch was dominated, newer platforms offer different benefits that might serve better specific learning goals. When looking into the best timing and pathways for transitions, this information would help shape curricular designs. When do students successfully move from block-based to text-based programing? what instructional approach support this shift the best? and what signs show a student is ready for more complex levels? Exploring the role of block-based programming in building wider computational thinking skills, would answer questions regarding transfer and general applicability,. Do students use this computational thinking outside of the programming context? If so, which educational methods, would maximize this transfer? Studying long-term results for students who started learning programming through block-based environments would give essential insights. Do these students continue in computer science more than those who start with text-based languages? And how do their later programming skills compare? Research about aspects of equity needs more in-depth exploration. Although this research finds encouraging results in gender equity, questions linger about socio-economic factors, racial and ethnic differences, and learning differences among students. How can block-based programming best serve diverse learners? Lastly, research should also look into block-based programming within diverse educational settings, not just school environments. Learning spaces outside school like afterschool programs, or family-based learning settings might offer different or additional methods for, developing computational skills.

  Block-Based Programming for Education: A Comprehensive Analysis of Visual Programming Environments in K-12 Learning · 2026 · DOI

Most-cited papers in Teaching and Learning Programming

• Integrating Ethics and Career Futures with Technical Learning to Promote AI Literacy for Middle School Students: An Exploratory Study · International Journal of Artificial Intelligence in Education · 2022 · 274 citations
• Artificial intelligence and human behavioral development: A perspective on new skills and competences acquisition for the educational context · Computers in Human Behavior · 2023 · 180 citations
• Learning to code and the acquisition of computational thinking by young children · Computers & Education · 2021 · 161 citations
• A systematic review of teaching and learning machine learning in K-12 education · Education and Information Technologies · 2022 · 158 citations
• A Revaluation of Computational Thinking in K–12 Education: Moving Toward Computational Literacies · Educational Researcher · 2021 · 152 citations
• AI + Ethics Curricula for Middle School Youth: Lessons Learned from Three Project-Based Curricula · International Journal of Artificial Intelligence in Education · 2022 · 151 citations
• Assessing systems thinking: A tool to measure complex reasoning through ill-structured problems · Thinking Skills and Creativity · 2018 · 145 citations
• Prompt Problems: A New Programming Exercise for the Generative AI Era · 2024 · 140 citations
• The Widening Gap: The Benefits and Harms of Generative AI for Novice Programmers · 2024 · 140 citations
• Understanding K–12 teachers’ technological pedagogical content knowledge readiness and attitudes toward artificial intelligence education · Education and Information Technologies · 2024 · 139 citations

Most recent work

• Generative Artificial Intelligence‐Assisted <scp>STEM</scp> Education: A Systematic Review and Future Research Agenda · European Journal of Education · 2026
• Integrating agile methodologies and AI-assisted learning in web programming education: a theoretical framework for CS curriculum transformation · Discover Education · 2026
• Designing an Interdisciplinary Artificial Intelligence Curriculum for Engineering: Evaluation and Insights From Experts · Computer Applications in Engineering Education · 2026
• A Systematic Review of Studies on the Use of Generative Artificial Intelligence Tools in Programming Education · Kastamonu Eğitim Dergisi · 2026
• CRAILF: A Zero-Cost Python-Based Gamified Framework for Enhancing AI Literacy Among Rural High School Students · Review of Artificial Intelligence in Education · 2026
• Rethinking sociology education in human–robot social interaction learning ecosystems: a mixed method study · International Journal of Sociology and Social Policy · 2026
• Digital Scaffolding as Learning Opportunities: Enhancing Vocabulary Knowledge Through Copilot AI · Educational Process International Journal · 2026
• Bibliometric Analysis of Computational Thinking Research in Vocational Education: Trends and Key Findings · TEM Journal · 2026
• Changes in Teacher Perceptions Through Professional Development on Integrating Generative AI into Geography Inquiry Activities · Journal of Geography · 2026
• South African Grade R teachers’ perspectives on integrating coding and robotics in early education · South African Journal of Education · 2026

Related topics in Computer Science