Quantifying Accuracy Improvement in Sets of Pooled Judgments
重新分析Herzog和Hertwig（2009）的数据，发现接受辩证指导语的参与者与接受标准指导语的参与者在两次估计平均后的准确性提升程度相同，质疑了辩证指导语能带来更大提升的结论。
Galton (1907) first demonstrated the "wisdom of crowds" phenomenon by averaging independent estimates of unknown quantities given by many individuals. Herzog and Hertwig (2009; hereafter H&H in Psychological Science) showed that individuals' own estimates can be improved by asking them to make two estimates at separate times and averaging them. H&H claimed to observe far greater improvement in accuracy when participants received "dialectical" instructions to consider why their first estimate might be wrong before making their second estimates than when they received standard instructions. We reanalyzed H&H's data using measures of accuracy that are unrelated to the frequency of identical first and second responses and found that participants in both conditions improved their accuracy to an equal degree.