Designers and Developers Grapple with LLM Integration as Tools Versus Teammates
A recent study available on arXiv (2604.15344v1) investigates how designers and developers from three leading tech firms incorporate large language models (LLMs) into their processes. Through interviews with 33 industry professionals, the findings indicate that the choice to utilize LLMs goes beyond mere technical aspects. Respondents often evaluated the potential function of an LLM within their workflows and its interaction with existing accountability and responsibility frameworks. When LLMs were viewed as tools under clear human oversight, their integration was generally welcomed and aligned with governance structures. However, perceiving LLMs as collaborators with unclear agency led to considerable reluctance, particularly regarding accountability for results. The study, using a constructivist grounded theory approach, emphasizes that these integration decisions are seldom straightforward. Participants also noted instances of effective collaboration with LLMs, reflecting complex and developing relationships with the technology.
Key facts
- Study published on arXiv with identifier 2604.15344v1.
- Research based on interviews with 33 designers and developers.
- Participants were from three large technology organizations.
- Study used a constructivist grounded theory methodology.
- LLMs framed as tools under human control were more readily accepted.
- LLMs framed as teammates with ambiguous agency caused hesitation.
- Decisions involved structures of responsibility and organizational accountability.
- Choices about LLM use are rarely purely technical or binary.
Entities
Institutions
- arXiv