How to design the interplay between humans and AI-based surface inspection systems

Computer assistance is indispensable for surface inspections in the steel industry, with artificial intelligence (AI) technologies commonly integrated into vision-based automated surface inspection systems (ASIS). However, the adoption of these technologies presents challenges in optimizing human-system interaction. The purpose of this research is to identify specific factors enhancing the acceptance of the system and to use these findings to formulate design principles for a user-centric ASIS. Semi-structured interviews were conducted with 12 stakeholders, including developers, operators, and quality staff, to understand their needs and preferences. Insights from these interviews led to the identification of seven essential design principles for advancing ASIS development. These principles include advanced user guidance, sorting and filtering options, a customer-specific knowledge base, and features that ensure consistently high-quality classifiers.