Analyzing Collaborative Challenges and Needs of UX Practitioners when Designing with AI/ML

UX designers and researchers who work with AI/ML face different kinds of challenges throughout the design process. Though close collaborations with AI/ML developers and data scientists could address some of these challenges, such interdisciplinary collaborations are non-routine and hard to realize. In this work, we investigate barriers for effective collaboration with ML practitioners, how they affect UX practice of AI/ML applications, and what UX practitioners need to overcome these challenges. We conducted a qualitative study with 14 UX practitioners who are working on AI/ML products as designers or researchers. Our findings show that UX practitioners face challenges in communication, understanding the model and model development processes, establishing ways to collaborate, and reconciling model-centric metrics of evaluation with user-centric outcomes. They described various needs in terms of more visibility into model development processes, access to comprehensible and contextual model information, and hypothetical tools that can potentially support collaboration with ML practitioners and enhance UX design processes. We discuss implications of this research for designing collaborative tools and empowering UX practitioners.