Recent advancements in AI have significantly enhanced collaboration between humans and writing assistants. However, empirical evidence is still lacking on how this collaboration unfolds in scientific writing, especially considering the variety of tools researchers can use nowadays. We conducted observations and retrospective interviews to investigate how 19 computer science researchers collaborated with intelligent writing assistants while working on their ongoing projects. We adopted a design-in-use lens to analyze the collected data, exploring how researchers adapt writing assistants during their use to overcome challenges and meet their specific needs and preferences. Our findings identify issues such as workflow disruptions and over-reliance on AI, and reveal five distinct design-in-use styles---teaching, resisting, repurposing, orchestrating, and complying---each consisting of different practices used by researchers. This study contributes to understanding the evolving landscape of human-AI co-writing in scientific research and offers insights for designing more effective writing assistants.
https://dl.acm.org/doi/10.1145/3706598.3713205
The ACM CHI Conference on Human Factors in Computing Systems (https://chi2025.acm.org/)