Fooled by Frequency—or Enlightened by It?

Humans tend to prefer a frequently sampled option over an infrequently sampled option even when the latter is more likely to result in a reward (i.e., frequency bias). To investigate the underlying mechanism of this “seemingly irrational” bias, we had participants (psychology undergraduate students) estimate the reward rate and their confidence in the estimate for both options. Across two experiments (N = 173), we found that 1) reward rate estimates were more accurate for frequently sampled options and 2) confidence in estimation accuracy was unaffected by the sampling frequency. Importantly, when frequently sampled options had objectively poor reward rates, participants shifted their preference toward the infrequently sampled option, suggesting that frequent sampling improved their ability to identify the better option. These results suggest that frequency bias reflects a “rational” use of frequently sampled information rather than an irrational distortion of reward-rate estimation or confidence.