The impact of distribution properties on sampling behavior

Area of Science:

• Cognitive Psychology
• Decision Science
• Behavioral Economics

Background:

• Individuals often overweight rare outcomes, influencing decisions and judgments.
• Previous research attributes this to cognitive biases like probability weighting flaws.
• This study investigates an alternative hypothesis: search behavior bias due to rare outcomes.

Purpose of the Study:

• To examine how rare outcomes in skewed distributions influence sampling behavior.
• To determine if increased sampling helps adjust estimations toward the true mean.
• To understand the cognitive processes underlying adjustments for sampling bias.

Main Methods:

• 180 participants completed an online shopping task involving product rating histograms.
• Participants sampled histogram bins representing customer reviews (positivity scores).
• Histogram shapes (skewness) and sample counts were manipulated to assess estimation accuracy.

Main Results:

• Binomial tests confirmed significant sampling bias toward rare outcomes (11-12% of samples).
• Estimates moved closer to the true mean with more samples, particularly in negatively skewed distributions (-7%).
• Despite adjustments, estimates deviated from the true mean by approximately 9.3%, showing persistent sampling bias.

Conclusions:

• Search biases systematically influence distributional judgments and estimation accuracy.
• Cognitive processes interact with biased sampling, affecting decision-making.
• Findings have implications for human-algorithm interactions in e-commerce, social media, and political contexts.