When does human-centered AI (fail to) scale?

State-of-the-art AI systems are built and deployed at the societal scale, increasing the need to consider sociotechnical factors for implementing systems of such magnitude. In contrast, individual user experience has long been the core of designing and developing user-friendly technologies. Through a series of experiments and case studies, I examine challenges and breakdowns as one extends individual-centered approaches to design societal-scale AI systems.