Abstract:

The spread of artificial intelligence (AI) in the modern digital world has posed new challenges for the field of higher education, particularly in the teaching of programming to university students. It is no secret that students use various AI tools in the learning process, including code development. However, how they do it, what tools they use, why they use them, and how this affects the quality of their programming education is a problem that requires research. In order to solve it, we have posed the following research questions:
1. What are the best practices for using AI in teaching programming, in particular when learning Python?
2. What is the state of students' use of AI tools when developing code?
3. Is it necessary and how to change (supplement) the policy of universities (technical and computer science faculties) on the use of AI, focusing on developing code?

To answer the first question, we analysed scientific articles and teachers' experiences to identify the best practices for using AI in teaching programming. These include guided code generation, AI-assisted debugging, formative feedback through conversational agents, and clear instructions for evaluating AI-generated code.

To answer the second research question, we surveyed students studying programming at universities in Belgium, Ukraine, Poland, Indonesia, and the United States. The survey is comprehensive and covers several aspects of AI use. The results show that most respondents use AI tools – primarily ChatGPT, GitHub Copilot, and online code assistants – to generate code snippets, debug errors, and clarify programming concepts. An important part of the survey is the questions of how students check the quality of the code created by AI and what precautions they take to avoid copying incorrect solutions. The survey also figures out the level of students’ awareness of university policies regulating the use of AI. Students demonstrate limited awareness of institutional AI policies and express a need for clearer guidelines. The ethical views of respondents are studied separately: from examples of honest use of AI to situations that they consider unethical. Students also expressed their own vision of how the use of AI should be regulated in programming courses. No less important for the study is the part of the survey that concerns the subjective feelings of students. We found out how they feel when using AI, whether it affects their confidence or programming skills, and what benefits they see in the process. Finally, the questionnaire asks students to evaluate how the use of AI has affected their grades, ability to complete tasks independently, and the speed of code development. While students report increased speed and confidence when coding, many acknowledge risks such as over‑reliance on AI and reduced independent problem‑solving.

In conclusion, based on the literature review and survey, the study proposes an update to university policies, including a clear definition of acceptable use of AI in coding, requirements for documenting AI use, the introduction of AI literacy training into programming courses, development of AI use policies for each programming and software development course, and assessments that take into account student work supported by AI and independent work. The authors provide recommendations for the development of responsible AI use policies in engineering and computer science departments.

Keywords:

Artificial intelligence, higher education, teaching programming, developing code, survey.