Empirical Evaluation of AI-Assisted Code Generation on Full-Stack Web Development Performance

The research performed a comparison study between manual development methods and AI-supported development methods for building complete web applications. The research team tested both development methods by creating the same functional modules to study how AI tools affected development speed and code quality and debugging activities and developer satisfaction. The study results demonstrate that AI-powered development speeds up projects because it automates standard tasks and produces complete development code. The system proves most effective during fast development and prototype creation for new products. The research demonstrates that AI-generated code needs testing and development work to achieve both logical accuracy and product-specific requirements. The manual development process enables developers to control every aspect of their code which leads to results that meet their specific needs. The process demands extra resources for execution yet it provides developers with complete system comprehension which helps them handle intricate situations better. The research results show that organizations must choose between two opposing goals which create a trade-off between operational efficiency and managerial authority. Human experts remain necessary because AI tools can boost work efficiency. The success of AI-assisted development work depends on three factors which include effective prompt engineering methods and developer supervision and the ability to assess output results. The study finds that AI technology functions as a supportive tool which enhances software development work without taking over developer responsibilities. Developers can achieve optimal results by using both manual coding methods and AI coding support which together create the most efficient solution. The research findings show how artificial intelligence technology changes software development methods. The research demonstrates how AI-assisted coding helps software development but also shows its limits which need to be addressed through a proper distribution between automated processes and human work in contemporary development environments.

[1] M. Chen, J. Tworek, H. Jun, et al., “Evaluating Large Language Models Trained on Code,” arXiv preprint arXiv:2107.03374, 2021. [2] E. Kalliamvakou, C. Bird, T. Zimmermann, et al., “The Impact of AI on Developer Productivity: Evidence from GitHub Copilot,” IEEE Software, vol. 41, no. 3, pp. 34–42, 2024. [3] H. Pearce, B. Ahmad, B. Tan, et al., “Asleep at the Keyboard? Assessing the Security of GitHub Copilot’s Code Contributions,” in IEEE Symposium on Security and Privacy, 2022, pp. 754–768. [4] P. Devanbu, T. Zimmermann, and C. Bird, “Beliefs and Evidence about Software Engineering,” in Proceedings of the 38th International Conference on Software Engineering, 2016, pp. 108–119. [5] S. Barke, M. B. James, and N. Polikarpova, “Grounded Copilot: How Programmers Interact with Code-Generating Models,” Proceedings of the ACM on Programming Languages, vol. 7, no. OOPSLA1, pp. 85–111, 2023. [6] P. Vaithilingam, T. Zhang, and E. L. Glassman, “Expectation vs. Experience: Evaluating the Usability of Code Generation Tools Powered by Large Language Models,” in CHI Conference on Human Factors in Computing Systems, 2024, pp. 1–23. [7] N. Perry, M. Srivastava, D. Kumar, and D. Boneh, “Do Users Write More Insecure Code with AI Assistants?” in ACM Conference on Computer and Communications Security, 2023, pp. 2785–2799. [8] N. Nguyen and S. Nadi, “An Empirical Evaluation of GitHub Copilot’s Code Suggestions,” in Proceedings of the 19th International Conference on Mining Software Repositories, 2022, pp. 1–5.