Our research on crowdsourcing aims to broaden the types of tasks that crowdsourcing systems can handle by developing new interfaces that enhance people’s ability to perform complex, unfamiliar, or ambiguous tasks. The premise is that we can achieve quality, efficiency or worker satisfaction outcomes by designing meaningful interactions between novices, experts and algorithms. We have developed crowdsourcing platforms that target scientific and medical tasks. In one of the projects, we developed CrowdEEG, a platform that enables medical specialists to collaboratively annotate sleep EEG data, and demonstrate that the deliberation process can improve output quality and efficiency. Our work is also increasingly concerned with understanding crowdwork practices, such as tooling practices and how they affect workers’ mental health and fair payment in collaborative tasks.