The article discusses the design of a tablet-based application called Co-ML meant to foster learning of Dataset Design Practices (DDPs) for creating inclusive ML models. The application allows beginners to build image classifiers in a distributed experience where data is synchronized across multiple devices. This promotes collaboration among users to refine ML datasets. The application was used in an AIML Summer Camp, where it helped students understand the importance of data diversity and quality in ML systems. The study found that students often prioritized learnability over class balance when improving model performance.
Publication date: 15 Nov 2021
Project Page: https://arxiv.org/abs/2311.09088v1
Paper: https://arxiv.org/pdf/2311.09088
