What should actually change in the classroom?
The core shift is from teaching coding as a solo craft to teaching coding as a human-AI collaboration. A 10-week case study with 25 engineering students found that using AI assistants like GitHub Copilot saved 30–40% of time on prototyping and debugging [1]. That time saving is huge—it means students can spend less time wrestling with syntax and more time on architecture, testing, and design decisions. But the same study found that maintaining code quality still required systematic human oversight [1], so the goal isn't to stop teaching coding; it's to teach students when and how to trust or override the AI.
Another study developed a structured AI-assisted learning flow for an engineering course and found that ChatGPT had significant potential to help students, but only if students were trained on how to interact with it effectively [2]. That means curricula need to include explicit instruction on prompt engineering, evaluating AI output, and knowing when the AI is likely wrong. Without that training, students may get generic or misleading responses [3].
What are the risks that teaching must address?
The biggest risk is that students become dependent on AI and stop learning the deep problem-solving skills that coding builds. In a study where GPT-3.5 was used as an AI tutor in an automated programming assessment system, students themselves raised concerns that using the AI tutor could inhibit their learning progress [3]. That's a red flag: if students feel the tool is making them dumber, educators need to design guardrails—like requiring manual coding first, then using AI for optimization or debugging.
There's also a serious legal risk that most courses ignore. A 2025 analysis warns that AI coding assistants can create intellectual property problems: the code they generate may violate open-source licenses or infringe third-party copyrights, which could prevent students from releasing their work as open source or even expose them to litigation [4]. Software engineering education must now include a module on the legal implications of AI-generated code, especially for students who plan to work in research or open-source projects.
Where do the studies agree and disagree?
All four studies agree on one central point: AI coding assistants are powerful tools that can boost productivity and satisfaction—96% of students in one study were satisfied with the AI-assisted learning experience [1]—but they require careful integration to avoid undermining learning. [1] and [2] both emphasize the need for structured pedagogical frameworks, while [3] adds the student perspective that overuse can backfire. [4] introduces a dimension the others miss: legal risk, which is a critical gap in current teaching.
The apparent conflict is about how much foundational coding still matters. [1] says human oversight is essential for code quality, implying that manual coding skills remain important. [3] shows students worry about learning inhibition, which reinforces that view. But [2] focuses more on how to use AI effectively, which could be read as de-emphasizing manual coding. The resolution is that both are right: students need to learn to code manually first, then learn to collaborate with AI—not skip straight to the AI.
About These Sources
This answer is built on 4 studies (2 peer-reviewed, 2 preprints) — published from 2023 to 2025, 3 from 2024 or later, 1 in Q1 journals, collectively cited 105 times — selected as the most relevant from 4 studies that passed quality screening, drawn from 28 papers retrieved from a database of over 500 million.
Sources used in this answer
Leveraging AI-Assisted Coding Tools in Engineering Education: Promise and Pitfalls in Software Development
In a 10-week case study with 25 engineering students building a CRM system, AI-assisted tools saved 30–40% of time on prototyping and debugging, but maintaining code quality required systematic human oversight; 96% of students reported satisfaction with the AI-assisted learning experience.
Digital transformation in engineering education: Exploring the potential of AI-assisted learning
In an engineering course at a Southeast Asian university, ChatGPT showed significant potential to assist students, but the study concluded that training and guidance on effective interaction with the AI tool are necessary for success.
AI-Tutoring in Software Engineering Education
In an exploratory case study integrating GPT-3.5 as an AI tutor in an automated programming assessment system, students identified advantages like timely feedback and scalability, but also raised concerns that the AI tutor could inhibit their learning progress.
Will AI coding assistants kill FLOSS in research software engineering?
A 2025 analysis warns that AI coding assistants pose intellectual property risks—including copyright infringement and open-source license violations—that could prevent research software engineers from releasing code under open-source licenses and expose them to litigation.