earticle

영어

This study explores the integration of Generative AI into Test-Driven Development (TDD) to efficiently produce code that accurately reflects programmers' requirements in software engineering. Using the Account class as an example, we analyzed the code generation capabilities of leading Generative AI models—OpenAI's ChatGPT, GitHub's Copilot, and Google's Gemini. Our findings indicate that while Generative AI can automatically generate code, it often fails to capture programmers' intent, potentially leading to functional errors or security vulnerabilities. By applying TDD principles and providing detailed test cases to the Generative AI, we demonstrated that the generated code more closely aligns with the programmer's intentions and successfully passes specified tests. This approach reduces the need for manual code reviews and enhances development efficiency. We propose a development process that combines TDD with Generative AI, leveraging the strengths of both to efficiently produce high-quality software. Future research will focus on extending this approach to more complex systems and exploring automatic test case generation techniques.