AI inom mjukvaruutveckling : Utvecklares användning av Large Language Models

Sammanfattning

Large Language Models (LLMs) have rapidly become a central component in software development, with the potential to fulfill various roles in day-to-day tasks. This has led to challenges for traditional professional roles and the emergence of new forms of learning, collaboration, and problem-solving. However, the impact of LLMs on software developers in this context is still relatively underexplored. This study therefore aims to investigate how language models influence the work of developers. Using a qualitative approach, consisting of semi-structured interviews with five professional software developers from two organizations, this study examines the types of learning that occur, how collaboration is affected, and the challenges that arise when integrating LLMs into work processes. The results indicate that developers learn both explicitly and implicitly through their interaction with the models. The ability to formulate clear and structured prompts significantly influences the models' ability to generate relevant responses, highlighting the importance of communication skills for effective use. Challenges related to the LLMs' limited context understanding and the need to incorporate larger codebases are also identified. Furthermore, the respondents become less reliant on collaboration with others, showing signs of role shifts in the field. This opens up opportunities for further research on how the use of LLMs affects roles and dynamics within cross-functional development teams