Collective decisions in social insects are made in self-organizing manner where group members only react to local information. This system exerts a great deal of power in a huge colony where a large number of individuals are performing enormous work. However, it is unknown whether decision making in self-organizing manner can be performed invariably under any social conditions. Here we show that the self-organization system is less efficient in the case with less work to do. When workers were given a lot of eggs in a Petri dish, they quickly gathered all the eggs into a single egg pile to tend them. The speed of egg piling increased as the number of workers. Interestingly, when workers were given only a few eggs, it took them much longer time to finish egg piling regardless of the number of workers. These results matched well with the prediction of our agent-based model with the algorithm that incorporated local communication mediated by a pheromone informing the location of egg pile. This study elucidated the major effects of the amount of work and the number of workers on the efficiency of self-organization system.