2024 · 19 citations · 14 references
ComputingComputer ScienceInformation TechnologyArtificial IntelligenceProgramming LanguagesSoftware EngineeringApplication Programming InterfacesArtificial Intelligence AdoptionChatbot AdoptionChatbotsCommercial ProgrammingSoftware Development Methodology
Programming assistants based on artificial intelligence (AI), such as ChatGPT and GitHub Copilot, have gained worldwide popularity recently. Studies in software development have explored the adoption of these tools, investigating their characteristics and impacts and how practitioners interact and perceive them. To contribute to this growing body of knowledge, in this study, we aim to explore the adoption of AI-based programming assistants in the Brazilian industry. More specifically, we aim to understand how practitioners of a particular Brazilian agroindustry-related company perceive and use AI-based tools to develop software. Using an online survey, we collected and analyzed 72 responses from employees of the studied company. Our findings suggest that practitioners mainly adopt ChatGPT and GitHub Copilot, interacting with these tools to accelerate online searching, typing, and syntax recall. A recurrent difficulty is the lack of context in the suggestions provided by these tools, but participants work on detailed descriptions to contextualize and cope with this challenge. Among the reasons for not using AI-based tools, the most influential is that participants use a commercial programming language, i.e., Uniface, which these tools lack examples. Our results provide insights into the state of the practice related to AI-based programming assistants and discuss implications for practitioners and researchers.
14
GitHub Copilot AI pair programmer: Asset or Liability?
Arghavan Moradi Dakhel, Vahid Majdinasab, Amin Nikanjam et al. · Journal of Systems and Software · 2023 · 340 citations
An empirical evaluation of GitHub copilot's code suggestions
Nhan Nguyen, Sarah Nadi · 2022 · 281 citations
Generative AI for Software Practitioners
Christof Ebert · IEEE Software · 2023 · 270 citations