Wisdom of the crowd is an intriguing phenomenon, which received its popularity in public discourse due to Galton when he surprisingly found a near-accurate estimate of the weight of a prize-winning ox by the common public. In other words, whatever errors in the guesses of the individuals got cancelled when averaged. An essential prerequisite of achieving the group’s wisdom is that the errors of individuals are statistically independent of each other.
Vul and Pashler took this to the next level. They wanted to test it for single individuals by allowing them to make multiple guesses. They asked 428 participants (through an internet poll) eight questions regarding general knowledge. Half of the participants were asked to make a second guess immediately after the questionnaire, and the other half were asked after three weeks (without telling them it would come after three weeks).
The results were surprising. The within-person average of the guesses (or the mean square error) was lower than either of the guesses for the immediate responders. The value was even lower for late-responders.
