Flatpack ML:How to support designers in creating a new generation of customizable machine learning applications

This paper examines how designers can be supported in creating a new genera-tion of interactive machine learning (ML) applications that run locally on con-sumer-level hardware and can be trained and customized by end-users for their specific context and use case. It delineates the proposed applications against re-search into Interactive Machine Learning and Machine Teaching, examines the challenges designers face in these contexts and their relevance for designing the proposed new applications, and reports on the findings of a survey exploring de-signers' interest in ML, their understanding of ML capabilities and concepts, and their preferences for learning about ML. Based on findings from the literature and the survey, the paper identifies three overlapping research challenges in supporting designers to ideate, design and prototype the proposed ML applications