The Utility of Theories in Intuitive Statistics: The Robustness of Theory-Based Judgments

Research on human judgment demonstrates that people's theories often bias their evaluation of evidence and suggest that people might be more accurate if they were unbiased by prior beliefs. Rather than comparing people's judgments of data when they do or do not have a prior theory, most studies compare people's estimates to conventional statistical standards, even though the status of these measures as normative criteria is controversial. We propose that people's theories may have beneficial consequences not examined in previous research. In two paradigms (the covariation estimation problem and the t-test problem), we compare judgments made by people who have potentially biasing prior information. We vary the quality of the data, presenting subjects with data that are either well-behaved or contaminated with outliers. In both paradigms, people's judgments approximated robust statistical measures rather than the conventional measures typically used as normative criteria. We find the usual biasing effects of prior beliefs but also find an advantage for subjects who have prior theories - even incorrect ones - over subjects who are completely "objective." Potentially biasing beliefs both enhanced people's sensitivity to the bulk of the data and reduced the influence atypical scores had on their estimates. Evidence is provided that this robustness results from the fact that prior theories make judgment problems more meaningful. We discuss the conditions under which prior beliefs are likely to help and hinder human judgment.