Figure 4.
The relationship between the average decision order during polling 2 for individuals who were wrong during the preceding polling 1 and the collective improvement (accuracy rate polling 2 – accuracy rate polling 1). Each dot represents one round. Collective improvements arose when individuals who were wrong during polling 1 decided late during the subsequent polling 2. When individuals who were wrong during polling 1 decided early, collective decisions could even become worse, though this happened relatively rarely.