Many activities in operations management are automated. Due to their flexibility, humans still play a critical role in many settings, and human-machine interactions need to be managed efficiently to ensure smooth operations. In the standard human-machine interaction at the shop-floor level, machines determine the assignment of tasks, while human workers mainly execute these often repetitive and monotonous tasks. One downside of such work division is the mental impoverishment of workers, which relates to stagnating productivity and low job satisfaction. We draw on goal-setting and self-determination theory to argue that enabling workers to self-select tasks can resolve these issues. We conduct a field study in a semi-automated warehouse to test if task self-selection improves workers' job performance and job satisfaction, compared to automatic task assignment by the machine. Indeed, we observe an 8.7\% increase in workers' job performance after the introduction of the task self-selection intervention; a great result considering the former year-long stagnating productivity. Contradicting our hypotheses, we also observe a decrease in job satisfaction, when "pickers could pick their picks". Triangulating surveys, focus interviews, and discussions with practitioners revealed that task self-selection suspended workers' prior informal work arrangements. A follow-up vignette study confirms that the intervention did not fulfill workers' need for self-determination, when these prior informal work arrangements were present. In their absence, we find a positive effect of the intervention on job satisfaction that is consistent with our hypotheses.
