AI in Computational Thinking Education in Higher Education: A Systematic Literature Review

AI summary

Overview

• Research area: Computing education / educational technology — specifically the intersection of Artificial Intelligence (AI) and Computational Thinking (CT) teaching in higher education. The paper is filed under AI Safety & Ethics (cs.CY) on arXiv.
• Technical level: Intermediate. No mathematics or model architectures are presented in depth, but the paper assumes familiarity with CT terminology (abstraction, decomposition, algorithms, debugging) and with broad AI categories (machine learning, deep learning, conversational AI, social robots).
• Scope: A systematic review of 18 peer-reviewed publications from 2017 to 2022, drawn from the ACM and IEEE digital libraries, that apply AI to CT education in higher education.

What This Paper Is About

Computational Thinking has been studied heavily in K-12 schooling but remains under-explored in higher education, leaving university teachers without a clear picture of what works. Meanwhile, AI is increasingly promoted as a way to support CT teaching, yet no consolidated overview existed of the benefits, challenges, teaching strategies, CT skills covered, or AI techniques used in this specific setting. The authors address this by conducting a systematic literature review of higher-education initiatives that apply AI to CT education.

Key Contributions

1. An inventory of AI-based CT education initiatives for higher education, compiled from a final set of 18 reviewed publications and organized into tables covering benefits, challenges, instructional strategies, CT components, and AI techniques.
2. A structured account of benefits and challenges. Benefits include AI's potential to reshape CT education, personalize learning, address diverse student CT backgrounds, build CT knowledge and skills, and motivate CT practice. Challenges include AI's potential to diminish students' creativity and innovation ability, the complexity of characterizing students' CT backgrounds and competencies, and the knowledge and expertise required to develop AI.
3. An expanded CT component framework. Using Brennan and Resnick's three dimensions (CT concepts, CT practices, CT perspectives) as a starting point, the review identifies a broader and partly fresh set of components, including new practices such as designing and constructing digital products, data organization/analysis, and reasoning, and an added perspective of computational empowerment drawn from Dindler et al.
4. Insights into new perspectives on CT in light of AI, notably the argument that the decoding approach should sit alongside the established coding approach, so students learn to examine and reason about the processes embodied in digital and AI-based systems rather than only writing programs.

The paper also notes that a poster based on this work was accepted and published in the Proceedings of the 30th ACM Conference on Innovation and Technology in Computer Science Education (ITiCSE 2025).

Main Findings

• The corpus is small. Running the search produced N = 64 publications for critical appraisal, which reduced to a definitive list of N = 18 publications selected for review.
• Generative AI was deliberately excluded. The review covers publications from 2017 to 2022 to avoid overlap with Generative AI and Large Language Model research, which the authors say began to proliferate from 2023 and which they excluded due to immature evidence of usefulness and lack of methodological consistency.
• Personalization is the most frequently cited benefit. Papers reported using recommendation systems, student profiling, neural network predictive models, and Bayesian hierarchical modeling to understand student profiles, address individual needs, predict struggles in advance, and implement adaptive activities.
• AI is reported to narrow the CT heterogeneity gap. AI-based courses such as CTQ (Computational Thinking Quest, an online adaptive gamified course) and adaptive games such as GidgetML are described as bridging differences in prior CT knowledge within a student group.
• Motivation and skill-building benefits appear across several papers. These include learning CT concepts in informal and public spaces such as museums (Forest Friends), solving contextualized real-world data-driven problems, supporting collaborative learning and extending CT learning beyond school walls, and providing creative learning experiences.
• Three challenges were identified. Diminishing students' creativity and innovation ability due to over-reliance on AI (reported by Han and Zhang); the complexity of characterizing students' CT background and competencies, because predictive student models often rest on simplifying or even invalid assumptions such as normal response distributions and homogeneous student characteristics; and the AI knowledge and expertise requirements for developing and sustaining AI-based initiatives.
• Five instructional strategies were observed: gamification, personalized/adaptive learning, project-based learning, task-based learning, and experiential learning.
• Concepts covered skew toward programming. The most frequently covered CT concepts were condition, iteration, and sequence; the perspectives dimension was the least covered, with only connecting and collaborating, and computational empowerment, identified.
• The AI techniques are diverse. Machine Learning, Deep Learning, Conversational AI, Data Science and analytics, Human-Computer Interaction, and Social Robots all appear across the reviewed work.

Methodology in Plain English

The authors followed the seven-step systematic literature review strategy suggested by Lockwood and Oh: developing structured questions, defining inclusion and exclusion criteria, developing a search strategy, critical appraisal, data extraction, analyzing extracted data, and presenting the findings.

• Inclusion criteria: peer-reviewed journal articles and conference publications, written in English, published between 2017 and 2022, focused on higher education, with full text available. Studies could be qualitative, quantitative, or mixed-methods, and could use AI in the educational design and intervention, the research methodology, or the evaluation.
• Search terms: for AI, "Artificial Intelligence", "AI", "Machine Learning", and "ML" combined with the boolean OR; for CT, "computational thinking"; for context, "University", "Universities", "Undergraduate", and "higher education". Searches were run in the ACM and IEEE digital libraries.
• Screening: two researchers — one with a background in software engineering and programming education, the other in AI and programming education — independently read titles and abstracts against four criteria: alignment with the search strategy's inclusion requirements, relevance to the research questions, use of AI in methodology or evaluation, and focus on computational thinking or programming as the core subject. Full texts were consulted when needed, and disagreements were resolved through discussion until consensus was reached.
• Analysis: an open thematic qualitative content analysis was used. For CT components, Brennan and Resnick's framework of CT concepts, practices, and perspectives served as the overall categories. For AI techniques, a bottom-up inductive approach was used with no pre-defined categories.

Why This Matters

Impact on research. The review situates CT education in higher education as an under-explored area and provides a consolidated baseline of 18 studies against which future work can be positioned. It also argues that CT's component set, long dominated by programming and coding, needs to be broadened toward "decoding" skills — understanding and reasoning about the processes inside digital and AI-based tools — which reframes a research agenda rather than just summarizing one.

Real-world applications:

• University instructors can use the inventory of tools and strategies (gamified courses, adaptive games, recommendation systems, chatbots, AR authoring tools) to select an approach suited to their course and student group.
• Curriculum designers can use the CT component table to check which concepts, practices, and perspectives their courses actually cover — particularly the under-represented perspectives dimension.
• Institutions planning AI-supported teaching can use the challenge list to anticipate the need for AI expertise, the risk of oversimplified student models, and the risk of students leaning on AI at the expense of their own creative thinking.
• Educators working with heterogeneous cohorts can look at the reported uses of AI to bridge gaps in students' prior CT and programming knowledge.

Industry relevance. The reviewed AI techniques include machine learning, deep learning with Artificial Neural Networks and Multi Feed-Forward architecture, convolutional neural networks, TensorFlow, conversational agents, AR, and social robots. This gives employers and technology providers a picture of the AI skills entering the graduate pipeline, and it flags the "decoding" capability — being able to interrogate how automated systems produce their outputs — as a workforce-relevant competency.

Future Directions

• Extend the timeframe beyond 2022 to capture Generative AI and Large Language Model research, which the authors deliberately excluded and which they say requires more time to establish consistency and reliable evidence.
• Broaden the search beyond ACM and IEEE to include Springer, Elsevier, Scopus, and Web of Science, as the authors explicitly recommend.
• Investigate whether and how AI shapes or replaces students' creative and computational thinking, including the mechanisms involved and the pedagogical implications — a call the authors note has also been made by other researchers.
• Develop the decoding and computational empowerment agenda further, examining how CT concepts and practices focused on understanding and reflecting on emerging digital technologies can be taught, and how to improve student models that currently rely on simplifying assumptions such as normal response distributions and homogeneous student characteristics.

Target Audience

Higher education instructors and curriculum designers who teach computational thinking, programming, or introductory computer science; computing education researchers looking for a structured baseline and open questions; and educational technologists or institutional decision-makers evaluating whether and how to introduce AI-supported tools into CT courses. Readers with no background in CT or AI will still follow the argument, but will get more from it with a basic grasp of both terms.

Authors’ abstract

Computational Thinking (CT) is a key skill set for students in higher education to thrive and adapt to an increasingly technology-driven future and workplace. While research on CT education has gained remarkable momentum in K12 over the past decade, it has remained under-explored in higher education, leaving higher education teachers with an insufficient overview, knowledge, and support regarding CT education. The proliferation and adoption of artificial intelligence (AI) by educational institutions have demonstrated promising potential to support instructional activities across many disciplines, including CT education. However, a comprehensive overview outlining the various aspects of integrating AI in CT education in higher education is lacking. To mitigate this gap, we conducted this systematic literature review study. The focus of our study is to identify initiatives applying AI in CT education within higher education and to explore various educational aspects of these initiatives, including the benefits and challenges of AI in CT education, instructional strategies employed, CT components covered, and AI techniques and models utilized. This study provides practical and scientific contributions to the CT education community, including an inventory of AI-based initiatives for CT education useful to educators, an overview of various aspects of integrating AI into CT education such as its benefits and challenges (e.g., AI potential to reshape CT education versus its potential to diminish students creativity) and insights into new and expanded perspectives on CT in light of AI (e.g., the decoding approach alongside the coding approach to CT).