Reaching for the star ratings: A Bayesian-inspired account of how people use consumer ratings

Online consumer ratings provide information about the average evaluation but also about the number of people who provided a rating (i.e., the sample size) and thus the rating's reliability. Using aggregated data, previous research has provided mixed results with regards to whether people pay attention to the sample size of statistical information. In our experiment, people choose between two competing hotels based on previous visitors' ratings. For each individual participant, we quantitatively estimated different strategies of how sample size and average ratings are processed. Results indicate substantial individual differences and show that people higher in statistical numeracy were better described by the Bayesian model. In addition, Bayesian strategies were used more when cues were incongruent (higher average rating has lower sample size) compared to when they were congruent. To account for these findings, we developed a Bayesian decision tree that describes in which situations behavior accords to Bayesian principles. Moving away from the debate about whether people are Bayesian or not, we explain who integrates sample size and average rating in a Bayesian way in which situation.