Crowdsourcing encompasses everything from large collaborative projects to microtasks performed in parallel and at scale. However, understanding subjective preferences can still be difficult: a majority of problems do not have validated questionnaires and pairwise comparisons do not scale, even with access to the crowd. Furthermore, in daily life we are used to expressing opinions as critiques (e.g. was cold, spicy, big), rather than describing precise preferences or choosing between (perhaps equally bad) discrete options. Unfortunately, is difficult to analyze such qualitative feedback, especially when we want to make quantitative decisions. In this article, we present collective criticism, a crowdsourcing approach where users provide feedback to microtasks in the form of critiques, such as it was easy/too challenging. This qualitative feedback is used to perform quantitative analysis of users' preferences and opinions. Collective criticism has several advantages over other approaches: too much/too little-style critiques are easy for users to provide and allows us to build predictive models for the optimal parameterization of the variables being critiqued. We present two case studies where we model: (i) aesthetic preferences in neural style transfer and (ii) hedonic experiences in the video game Tetris. These studies demonstrate the flexibility of our approach, and show that produces robust results that are straightforward for experimenters to interpret and inline with users' stated preferences.