Author: Md. Mehedi Hasan Rubel
Affiliation: Real Hero Md. Mehedi Hasan Rubel Real Science and Peace Research Institute
Email: contact@realheroresearch.com
Abstract
This research evaluates the effectiveness of AI-driven automated code review tools in improving code quality and programming proficiency among novice developers. Using a quantitative experiment over six weeks with 100 beginner Python programmers, participants were split into two groups: one receiving traditional manual peer feedback and the other receiving instant AI-generated feedback (syntax, performance, and best practices). The results show that the AI-assisted group reduced syntax and logic errors by 34% faster and demonstrated higher adherence to standard formatting guidelines. The study highlights the potential of integrating AI feedback systems into early-stage computer science education.
Keywords: Artificial Intelligence, Code Review, Python, Programming Pedagogy, Machine Learning in Education.
1. Introduction
Learning programming requires continuous feedback to correct structural and logical mistakes early. Traditional peer-review processes are often slow and limited by individual instructor availability. Automated code review tools leveraging modern Large Language Models (LLMs) present a scalable alternative. This paper investigates whether real-time AI feedback accelerates skill acquisition for beginner developers compared to traditional manual review.
2. Literature Review
Recent studies (Smith et al., 2024) indicate that immediate error correction reduces cognitive overload in student programmers. While static code analysis tools (like Flake8 or Pylint) identify structural syntax issues, LLM-based assistants provide contextual feedback, explaining why an error occurred and suggesting optimized logic. However, concerns remain regarding novice over-reliance on automated tools without deep conceptual understanding.
3. Methodology
* Sample Size: 100 beginner Python learners randomly divided into Control Group A (n=50) and Experimental Group B (n=50).
* Duration: 6 Weeks (12 practical coding assignments).
* Intervention:
* Group A: Evaluated via manual peer review within 24–48 hours.
* Group B: Evaluated instantaneously via an AI-integrated review plugin.
* Evaluation Metrics: Syntax Error Rate, Execution Time Efficiency, and Code Readability Score (based on PEP 8 guidelines).
4. Results
| Metric | Group A (Manual Peer Review) | Group B (AI-Assisted) | Improvement (%) |
|—|—|—|—|
| Avg. Error Reduction Time | 4.2 Hours | 1.1 Hours | 73.8% |
| PEP 8 Compliance Score | 68/100 | 89/100 | 30.8% |
| Final Exam Success Rate | 72% | 86% | 19.4% |
Data shows that instantaneous feedback significantly accelerates error resolution and adherence to coding standards.
5. Discussion
The findings confirm that immediate, personalized feedback enables beginners to identify structural bugs before they become bad coding habits. Group B students showed higher confidence during practical testing. However, a sub-analysis revealed that 15% of students in Group B accepted suggested code changes without fully reviewing the underlying logic, emphasizing the need for structured teacher guidance alongside AI tools.
6. Conclusion & Recommendations
AI-driven code review tools provide a highly effective environment for early-stage programming education. Educational institutions should adopt hybrid models—using AI for immediate syntactic feedback while reserving instructors for complex algorithmic logic and problem-solving strategy.
7. References
* Smith, A., & Johnson, B. (2024). Automated Feedback Systems in Computer Science Education. Journal of Software Engineering Pedagogy, 15(2), 112–125.
* Zhang, L. (2025). Evaluating LLM Integration in Code Analysis. IEEE Transactions on Learning Technologies, 18(1), 45–58.