What If We Ask A Different Question?: Social Inferences Create Product Ratings Faster

摘要

Consumer product reviews are the backbone of commerce online. Most commonly, sites ask users for their personal opinions on a product or service. I conjecture, however, that this traditional method of eliciting reviews often invites idiosyncratic viewpoints. In this paper, I present a statistical study examining the differences between traditionally elicited product ratings (i.e., "How do you rate this product?") and social inference ratings (i.e., "How do you think other people will rate this product?"). In 5 of 6 trials, I find that social inference ratings produce the same aggregate product rating as the one produced via traditionally elicited ratings. In all cases, however, social inferences yield less variance. This is significant because using social inference ratings 1) therefore converges on the true aggregate product rating faster, and 2) is a cheap design intervention on the part of existing sites.