Designing AI-Assisted Writing Interfaces for Requirements Engineering - Exploring Factors Influencing Usability and Design Guidelines

Abstract

The integration of Large Language Models (LLMs) into professional workflows has created new opportunities for supporting information-intensive tasks such as Requirements Engineering (RE). However, despite the growing interest in AI-assisted writing tools, limited research has explored how such tools should be designed to support usability in industrial RE contexts. This thesis investigates how interaction design can support the integration of AI-assisted writing tools into requirements management systems to assist engineers in formulating clear and consistent require ments. The study was conducted in collaboration with Geely Technology Europe and focused on a custom-built extension for the requirements management platform SystemWeaver. Using a Research through Design approach combined with the Double Diamond framework, the project included a literature review, semi-structured interviews, thematic analysis, ideation, prototyping, and usability testing. The results identified four key factors important for usability when integrating AI assisted writing tools in requirement management systems: control, trust, flexibility, and efficiency. Based on these findings, fourteen design guidelines across interaction, functionality, and design were developed The study highlights the importance of balancing trust and over-reliance, efficiency and control, as well as flexibility and consistency when integrating AI into professional requirements engineering environments. The findings contribute to the fields of Human-Computer Interaction, Human-AI Interaction, and Requirements Engineering by providing practical design guidance for future AI-assisted writing tools in industrial contexts.