Figure 1.

(a) Design of experimental tasks. As the internally guided decision making (IDM) and externally guided decision making (EDM) tasks, we used occupation preference judgment¹²,³⁴, for which no objectively correct answer exists, and salary judgment for which one objectively correct answer exists, respectively. Participants performed counterbalanced tasks of two types. RT denotes the reaction time. (b) Pre-rating–decision consistency (i.e., how often participants’ decisions of preference judgment were consistent with pre-ratings of the occupational preference), and average annual salary database–decision consistency (i.e., how often participants’ decisions were consistent with the actual average annual salary, which is based on a statistical survey by the Ministry of Health, Labour and Welfare of Japan) for each decision-making task (see Supplemental Information for more detailed methods to calculate these indexes). ***Denotes a significant main effect of epoch (p < 0.0001). Error bars show standard errors. (c) Schematic figure of the calculation of decision consistency score. The decision consistency score represents the rate of trials in which a certain occupation word was chosen or rejected repeatedly. In cases where a participant chose A (first trial in the example of this figure) and it was chosen again in the trial in which A was presented the next time (third trial in this example), we counted that trial as a consistent decision. In addition, in cases where a participant rejected B (second trial in this example) and it was rejected again in the trial in which B was presented the next time (fifth trial in this example), we counted that trial as a consistent decision. The consistent decisions were counted for each occupation word. Then that number was converted to a rate of consistent decision by dividing the number of consistent decisions by the sum of the number of consistent and inconsistent decisions. The average of the rate of consistent decisions was calculated across all occupation words. (d) Decision consistency scores for the first-half and the last-half trials in IDM (preference judgment) and EDM (salary judgment) tasks. *Denotes a significant main effect of epoch (p < 0.01). Error bars show standard errors.