work: Social mechanisms underlying helpfulness
evaluation. In this paper, we formulate and assess a set of theories that govern the evaluation of opinions, and apply these to a
dataset consisting of over four million reviews of roughly 675,000
books on Amazon’s U.S. site, as well as smaller but comparablysized corpora from Amazon’s U.K., Germany, and Japan sites. The
resulting analysis provides a way to distinguish among competing
hypotheses for the social feedback mechanisms at work in the evaluation of Amazon reviews: we offer evidence against certain of
these mechanisms, and show how a simple model can directly account for a relatively complex dependence of helpfulness on review and group characteristics. We also use a novel experimental
methodology that takes advantage of the phenomenon of review
“plagiarism” to control for the text content of the reviews, enabling
us to focus exclusively on factors outside the text that affect helpfulness evaluation.