People often conduct preprocessing to simplify primary processing. However, there is a trade-off between the costs of preprocessing and primary processing. Therefore, it is considered that the utility of preprocessing differs depending on the task complexity. We conducted five experiments to find out whether people could adaptively estimate the utility of preprocessing and take rational action when there is a trade-off between the costs of preprocessing and primary processing. In these experiments, we used a routine task of transcribing scores from fifty individual test sheets to a tally sheet to correspond to each student ID number. In this task, preprocessing is preliminarily grouping the test sheets on a desk according to the student ID numbers. Primary processing is transcribing the scores on the test sheets to the tally sheet to correspond to the student ID numbers. We set up high and low complexity conditions in two different ways of matching the student ID numbers. In Experiment 1, the participants were instructed to conduct or not to conduct preprocessing. As a result, in the high complexity condition, the task completion time was faster and the number of errors was smaller when preprocessing was conducted. On the other hand, in the low complexity condition, the task completion time was faster when preprocessing was not conducted, and the number of errors did not differ when preprocessing was conducted and when it was not. From these results, it is revealed that in the high complexity task, conducting preprocessing is effective, and contrarily, in the low complexity task, not conducting preprocessing is effective. In Experiment 2, using the same task as in Experiment 1, the participants were allowed to choose whether to conduct preprocessing or not. As a result, in the high complexity condition, almost all participants conducted preprocessing. In the low complexity condition, the participants learned to conduct preprocessing despite it not being effective. Experiment 3 was conducted in the same manner as in Experiment 1, using a smaller desk for preprocessing which makes conducting preprocessing more difficult. As a result, however, the participants managed to conduct preprocessing and showed the same performance as they did in Experiment 1. Experiment 4 was conducted in the same manner as in Experiment 2 using the smaller desk. As a result, in both high and low complexity conditions, almost all participants conducted preprocessing despite it not being effective in the low complexity condition. In Experiment 5, we investigated how the utility of preprocessing was estimated in the low complexity condition. As a result, we found that the participants overestimated the utility of preprocessing and chose to conduct preprocessing.