ARTIFICIAL INTELLIGENCE-BASED ADAPTIVE TEACHING OF PROGRAMMING IN THE DIGITAL EDUCATION ENVIRONMENT

УДК 004.9+371.3

Abstract

The article examines the scientific, theoretical and methodological foundations of artificial intelligence-based adaptive teaching of programming in the digital education environment. The relevance of the study is determined by the heterogeneity of students’ initial training, differences in algorithmic thinking, the need for rapid diagnostic feedback and the methodological limitations of traditional learning management systems. The paper substantiates that programming should not be taught in a digital environment as the simple delivery of electronic content, but as a data-informed pedagogical process that continuously identifies the learner’s current state, adapts the difficulty of tasks, and supports individual learning trajectories. The study systematizes the didactic functions of machine learning, learning analytics, knowledge tracing, recommender systems, intelligent tutoring systems, generative AI and automated feedback in programming education. A learner model, an error profile and a human-in-the-loop adaptive mechanism are proposed as key components of the methodology. It is concluded that the effectiveness of AI-based adaptive teaching depends not only on technological tools, but also on their integration with didactic objectives, formative assessment, academic integrity and teacher supervision.

Аннотация

В статье рассматриваются научно-теоретические и методические основы адаптивного обучения программированию на основе искусственного интеллекта в цифровой образовательной среде. Актуальность исследования обусловлена различием начальной подготовки студентов, уровней алгоритмического мышления, потребностью в оперативной диагностической обратной связи и методическими ограничениями традиционных LMS-платформ. Обосновывается, что обучение программированию в цифровой среде должно рассматриваться не как простое размещение электронного контента, а как педагогически управляемый процесс, основанный на анализе учебного состояния студента, адаптации сложности заданий и формировании индивидуальной образовательной траектории. Систематизированы дидактические функции машинного обучения, learning analytics, knowledge tracing, рекомендательных систем, интеллектуальных тьюторских систем, генеративного ИИ и автоматизированной обратной связи. В качестве ключевых компонентов методики предложены модель студента, профиль ошибок и адаптивный механизм с обязательным участием преподавателя. Сделан вывод, что эффективность адаптивного обучения на основе ИИ определяется не только технологическими возможностями, но и их согласованностью с дидактическими целями, формативным оцениванием, академической честностью и педагогическим контролем.

Keywords: artificial intelligence; adaptive teaching; programming education; digital education; learning analytics; individual learning trajectory; automated feedback.

Ключевые слова: искусственный интеллект; адаптивное обучение; обучение программированию; цифровое образование; learning analytics; индивидуальная образовательная траектория; автоматизированная обратная связь.

Introduction

The rapid development of digital education requires higher education institutions to reconsider the content, forms and methodological support of teaching. This requirement is especially evident in programming courses, where students enter the learning process with different levels of prior knowledge, different rates of mastering concepts, and different abilities to interpret algorithms, trace code and detect errors. Therefore, the digitalization of programming education cannot be reduced to posting lectures, tests and assignments on a platform. It should be organized as a flexible pedagogical system in which the learner's current state is continuously diagnosed and the learning process is adapted on the basis of reliable educational data.

Traditional LMS platforms provide essential organizational infrastructure: they store learning materials, collect assignments, administer tests, record grades and enable communication. However, their standard use often remains insufficient for deep personalization. They may show the final score, but they usually do not explain the type of programming error, the sequence of failed attempts, the learner's strategy of problem solving or the weak links between topics. This limitation makes adaptive teaching especially important for programming education.

Artificial intelligence technologies open new possibilities for such adaptation. Machine learning models can predict difficulties, learning analytics can reveal activity patterns, knowledge tracing can track the acquisition of specific programming concepts, recommender systems can select relevant resources and generative AI can provide explanatory feedback. At the same time, the use of AI in education requires pedagogical supervision, transparency, data protection and a balance between automation and the development of independent thinking.

Materials and Methods

The study is based on theoretical analysis, comparative generalization, pedagogical modeling and systematic interpretation of contemporary research on adaptive learning, intelligent tutoring systems, learning analytics, educational data mining and generative AI in programming education. The methodological focus is placed on the relationship between digital indicators of learning activity and didactic decisions made by the teacher or by an AI-supported adaptive system.

The source material was transformed from a dissertation chapter devoted to the scientific and methodological foundations of AI-based adaptive teaching in digital education. The chapter content was adapted to the structure of a journal article by narrowing the scope, removing dissertation-style subdivisions, formulating the aim and methods explicitly, and presenting the main results through a compact table and an integrated adaptive mechanism.

For the purposes of the article, programming education is interpreted as a process in which conceptual knowledge, procedural skills and creative problem solving develop simultaneously. This interpretation makes it possible to connect syntactic, semantic, runtime and algorithmic errors with a learner model and with the selection of adaptive tasks, explanations and feedback.

Results and Discussion

The analysis shows that the main didactic problem of programming education in a digital environment is the mismatch between a uniform learning path and the diversity of students' learning states. The same topic, the same task and the same assessment criteria do not have the same educational value for all students. A learner who has mastered conditional statements but struggles with loops needs visual algorithmic explanations and gradual practice. A stronger learner may need complex problem-based tasks, refactoring, optimization or project components.

In this context, adaptive teaching can be defined as a dynamic didactic system that adjusts learning content, task complexity, feedback format and repetition frequency according to the learner's current state. The central element of this system is the learner model. In programming education, the model should include not only test scores, but also the level of algorithmic thinking, the pace of task completion, the number of attempts, the types of code errors, the quality of explanation and the response to feedback.

A simplified representation of the learner model can be expressed as follows: L = {K, E, P, A, F}, where K denotes the knowledge state, E denotes the error profile, P denotes the learning pace, A denotes activity indicators, and F denotes the learner's response to feedback. The adaptive decision can then be represented as R = f(L, C, G), where C is the content model and G is the learning goal. This formulation emphasizes that adaptation is not a random recommendation, but a pedagogically justified decision based on the relationship between the learner's state, the structure of the content and the intended competence.

Table 1. Pedagogical problems and adaptive AI-based solutions in programming education

Table 1 demonstrates that each programming difficulty has a measurable digital indicator and can be connected with an adaptive pedagogical decision. In such a system, artificial intelligence does not replace the teacher; it transforms raw digital traces into diagnostic information and supports the teacher's decision-making. This distinction is essential because the same numerical indicator may have different pedagogical meanings. For example, many attempts may indicate insufficient understanding, but they may also show persistence and independent exploration.

The didactic functions of AI in adaptive programming education can be grouped into five categories. The first function is diagnostic: AI helps identify the learner's current state and the causes of errors. The second function is predictive: AI can reveal students who are likely to experience difficulties in specific topics such as loops, functions or arrays. The third function is adaptive: AI supports the selection of tasks, examples and resources that correspond to the learner's needs. The fourth function is feedback: AI can explain the reason for an error and suggest a correction strategy. The fifth function is analytical support for the teacher: dashboards and reports can show which topics cause the greatest difficulties for the group.

Generative AI has particular value in programming education because it can provide natural-language explanations of code, generate test cases, propose alternative solutions and answer student questions in dialogue form. Nevertheless, its use must be controlled methodologically. If the learner receives ready-made code without explaining its logic, the visible result may improve while algorithmic thinking remains undeveloped. Therefore, assignments should include explanation, reflection, code defense and process assessment.

Didactic mechanism of AI-based adaptive teaching. The proposed didactic mechanism consists of five consecutive and cyclical stages. At the first stage, the system conducts an initial diagnosis of prior knowledge and programming skills. At the second stage, the learner model is formed and updated using test results, code submissions, error types, time indicators and activity logs. At the third stage, adaptive decisions are made: the system recommends a resource, task, explanation, hint or repetition block. At the fourth stage, the student performs the task and receives individualized feedback. At the fifth stage, the result is interpreted by the teacher, and the learner model is updated for the next cycle.

The mechanism is based on two methodological principles. The first principle is continuity: diagnosis is not performed once, but throughout the learning process. The second principle is pedagogical control: AI recommendations are not final decisions, but information that supports the teacher's professional judgment. This approach preserves the educational role of the teacher and prevents the adaptive system from turning into an uncontrolled algorithmic environment.

The mechanism also requires ethical safeguards. The learner must understand how AI is used, what data are analyzed, and how recommendations affect learning tasks. Assessment should be transparent and should include both the final program and the learner's explanation of the algorithm. Such requirements reduce the risks of academic dishonesty, overreliance on generative AI and unfair algorithmic recommendations.

Conclusion

The study confirms that AI-based adaptive teaching of programming is a necessary methodological direction in the digital education environment. Programming courses require personalization because students differ in prior knowledge, algorithmic thinking, pace of learning, error patterns and motivation. A uniform digital course cannot fully respond to these differences.

Traditional LMS platforms provide important infrastructure, but they are not sufficient for dynamic methodological decision-making. Effective adaptive teaching requires a learner model, an error profile, a content model, adaptive tasks and individualized feedback. Artificial intelligence technologies can support these elements through diagnostics, prediction, recommendation, automated feedback and learning analytics.

At the same time, AI should be interpreted as a didactic mechanism that supports the teacher rather than a tool that replaces pedagogical judgment. The effectiveness of adaptive teaching depends on the integration of technological possibilities with learning objectives, formative assessment, academic integrity, transparency and human-centered pedagogical control. These findings provide a theoretical basis for designing an AI-supported adaptive methodology for teaching programming in higher education.

References:

1. Brusilovsky P. Adaptive hypermedia // User Modeling and User-Adapted Interaction. – 2001. – P. 87–110.
2. Corbett A.T., Anderson J.R. Knowledge tracing: Modeling the acquisition of procedural knowledge // User Modeling and User-Adapted Interaction. – 1995. – P. 253–278.
3. Woolf B.P. Building Intelligent Interactive Tutors: Student-centered strategies for revolutionizing e-learning. – Burlington: Morgan Kaufmann, 2009
4. Koedinger K.R., Corbett A.T., Perfetti C. The Knowledge-Learning-Instruction framework: Bridging the science-practice chasm to enhance robust student learning // Cognitive Science. – 2012. – P. 757–798.
5. VanLehn K. VanLehn K. The relative effectiveness of human tutoring, intelligent tutoring systems, and other tutoring systems // Educational Psychologist. – 2011. – Т. Vol. 46(4). – P. 197–221.
6. Siemens G. Learning analytics: The emergence of a discipline // American Behavioral Scientist. – 2013. – P. 1380–1400.
7. Ferguson R. Learning analytics: drivers, developments and challenges // International Journal of Technology Enhanced Learning. – 2012. – P. 304–317.
8. Romero C., Ventura S. Educational data mining and learning analytics: An updated survey // Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery. – 2020.
9. Sweller J. van Merrienboer J.J.G., Paas F. Cognitive architecture and instructional design: 20 years later // Educational Psychology Review. – 2019. – P. 261–292.
10. Kodirov, B. (2026). O'quvchilarda kompleks ma'lumotlar bazasi kompetensiyalarini rivojlantirish uchun tizimlangan pedagogik strategiyalar. Osiyo ilmiy tadqiqotlar va innovatsiyalar jurnali, 5 (1), 143 – 147.
11. Bobur, K. (2026). Ma'lumotlarni tahlil qilishda kompetentlikga asoslangan yundashining ahamiyati. Universum: psixologiya va obrazovanie, 2 (1 (139)), 33 – 37.
12. Akmalova, A., Xaydarov, A., & Pardayeva, G. (2024). Oliy ta’limni raqamlashtirish sharoitida talabalarning intellektual salohiyatini rivojlantirish. «ACTA NUUz», 1(1.1. 1), 40 – 43.
13. Abdunazarovna P.G. (2026). A project-based way to use open biological data in bioinformatics education. Universum: психология и образование, 2(1 (139)), 31 – 32.
14. Ugli N.B.U., & Ugli A.T.K. (2026). Adaptive multimodal biometric authentication systems: a human-centered analysis of design, evaluation, and security challenges. Universum: технические науки, 6(1 (142)), 69 – 72.
15. Rashidovich B.G., Ugli N.B.U., & Ugli N.B.B. (2026). Using ai to make a sophisticated decision-making system that can advise you in real time whether to stop or keep going at pedestrian crossings. Universum: технические науки, 6(1 (142)), 50 – 54.
16. Ugli, N.B. U., & Bakhtiyorovna, N.S. (2026). Advanced biometric authentication systems: an in-depth study of algorithmic models, security threats, and multimodal evaluation frameworks. Universum: технические науки, 6(1 (142)), 64 – 68.