Publication
Addressing the Cold-Start Problem in Case-Based Design Decision Support: Requirements for AI-Assisted Product Development
Kokulan Thanabalan; Luca Lorenz; Leonhard Kunz; Lukas Malburg; Ralph Bergmann
In: Johannes Wichmann; Lisa Grewenig; Pascal Reuß (Hrsg.). Workshop SIG Knowledge Management (FG WM), Workshop Proceedings. Workshop SIG Knowledge Management (FG WM-2026), located at KI 2026, August 12, Bremen, Germany, CEUR Workshop Proceedings, CEUR-WS.org, 2026.
Abstract
Experience-based design decision support systems can help product developers reuse knowledge from previous projects, avoid repeated mistakes, and identify suitable solution strategies for new design problems. Case-Based Reasoning is a promising approach for this purpose, as it supports recommendations based on similar past design situations. However, such systems face a cold-start problem when only few or no relevant cases are available, limiting their usefulness during early deployment or in data-sparse design contexts. This paper investigates how case-based design decision support can be extended to cold-start situations through AI-assisted support. Following a Design Science Research approach, an exploratory survey with product development experts was conducted to identify expectations, needs, and concerns regarding AI-generated recommendations in the absence of sufficient historical experience. The results show that users expect CAD-integrated support, multiple alternative solution suggestions, clear distinctions between retrieved and synthesized recommendations, explainable reasoning, confidence or uncertainty indications, and lightweight feedback mechanisms. Based on these findings, the paper derives requirements for AI-assisted cold-start support in case-based design decision support systems. The requirements provide a conceptual foundation for systems that support designers in data-sparse situations while gradually building reusable experiential knowledge through user feedback and documented decision outcomes.