Scholarlink

In this study, we investigate how developers interact with ChatGPT when seeking programming assistance to identify trends in language preferences and access communication efficiency. We analyze a public dataset, DevGPT, comprising 17,622 ChatGPT prompts, and 12,031 code snippets. Our findings show Python, Bash, and JavaScript as primary languages used in developer-ChatGPT interactions. We find that data manipulation/exchange languages/formats, namely JSON and SQL, require fewer interactions, while general-purpose languages such as C++ and C# need lengthier conversations. The findings provide valuable insights for improving training programs and enhancing ChatGPT’s language understanding, ultimately optimizing collaboration between developers and AI tools in software development.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Analyzing ChatGPT Assistance in Programming

  • 章节

  • 发表日期：

    2025年

查看全文
  • Costain Nachuma,
  • Md. Fazle Rabbi,
  • Arifa I. Champa,
  • Minhaz F. Zibran

摘要

In this study, we investigate how developers interact with ChatGPT when seeking programming assistance to identify trends in language preferences and access communication efficiency. We analyze a public dataset, DevGPT, comprising 17,622 ChatGPT prompts, and 12,031 code snippets. Our findings show Python, Bash, and JavaScript as primary languages used in developer-ChatGPT interactions. We find that data manipulation/exchange languages/formats, namely JSON and SQL, require fewer interactions, while general-purpose languages such as C++ and C# need lengthier conversations. The findings provide valuable insights for improving training programs and enhancing ChatGPT’s language understanding, ultimately optimizing collaboration between developers and AI tools in software development.