Abstract
Various information are taken into account on decision making. For example, we determine the transportation means or outfit of the day based on “weather forecast”. In addition, the behavior of a person also depend on the reliability of the information. In this study, we consider the case of repeating the action of selecting one from multiple choices with unknown rewards, which is well known task as multi-armed bandit, in pre-presented information. At the first time, pre-presented information is unreliable, and the effect of the information to the behavior is small. Participants can easily modify information to match the results of task if the information should be incorrect. We assume that participants may become difficult to correct the behavior due to prior information after correct information continues to be presented. It is because the reliability of information increases by repetitive correct information. Therefore, we verified how the information is corrected by the actual results and how the correction could be modified by the reliability of the pre-presented information. In particular, we analysis the participant behavior in the situation that the incorrect information is suddenly presented after the correct information is continuously presented. As a result, we found that the ratio of the incorrect choice is increased and the fixation of the behavior is delayed after the repetitive correct information.
