The Many Chefs in the AI Kitchen: Collaborative AI Design and Development in Oncology

Advancements in artificial intelligence (AI) are redefining innovation in healthcare. In the radiology domain, advanced AI systems have shown promising performance in medical tasks such as tumor detection and treatment plan. Given the multidisciplinary nature of the medical AI development team, challenges in collaboration, design, and integration are common. For example, the design lens is less present, researchers from different disciplines speak different languages, and clinicians as the end-users often play the external consultant role rather than team members. To address these gaps, we present the development of a collaboration platform, an AI-assisted computer-aided diagnosis (CAD) system for radiology, using a Research through Design (RtD) methodology. Through a three-phase case study, we showcase a novel way of operationalizing multidisciplinary collaboration with design team playing the coordinating and integrating role, highlighting the importance of including subject-matter experts directly in development teams. We also provide a design and development guideline for multidisciplinary research team. Moreover, this case study demonstrates how design struggles and decisions contribute to knowledge generation, offering strategies to balance human- and techno-centric perspectives. Reflecting on the process, we provide pertinent input for design research, HCI, and design and collaboration implications for developing AI systems in radiology, and for healthcare AI more generally.