We investigate the laws governing people's decisions and interactions by studying the collective dynamics of a well-documented social activity for which there exist ample records of the perceived quality: the attendance to movie theaters in the US. We picture the flows of attendance as impulses or "shocks" driven by external factors that in turn can create new cascades of attendances through direct recommendations whose effectiveness depends on the perceived quality of the movies. This corresponds to an epidemic branching model comprised of a decaying exponential function determining the time between cause and action, and a cascade of actions triggered by previous ones. We find that the vast majority of the ~3,500 movies studied fit our model remarkably well. From our results, we are able to translate a subjective concept such as movie quality into a probability of the deriving individual activity, and from it we build concrete quantitative predictions. Our analysis opens up the possibility of understanding other collective dynamics for which the perceived quality or appeal of an action is also known.