Trustworthiness of online beer ratings as a source of social information

摘要

People increasingly use the internet as a source of social information. The pay-offs associated with using such information depend on its quality in terms of content and bias. A key question is therefore what information is contained in socially acquired information and whether it is biased. For example, social information may be biased due to conformism if people 'producing' (i.e. posting) information adjust it based on existing social information. We addressed these questions by focussing on ratings of beers posted by Finns on the internet, which people use as a source of social information when making consumer decisions. To model the information contained in beer ratings, we analysed a repeated measures longitudinal dataset of >130 000 beer ratings collected by 490 Finns and estimated key variance components. We decomposed variation in social information (i.e. ratings) into variation attributable to characteristics of the beer (beer identity, beer style, brewery and country of brewery), characteristics of the (individual) rater, variation caused by temporal effects and residual variation. Moreover, we compared blind with non-blind rating scores to evaluate whether conformism represented a source of bias. The majority (65.1%) of the variation in beer ratings was explained by beer characteristics, 9.5% by the identity of the rater and 0.5% by temporal effects; only 25.1% of the variance remained unexplained. Blind ratings were positively correlated with non-blind ratings, suggesting that conformism did not introduce a major bias. Our findings imply that beer ratings posted on the internet may represent a relatively unbiased and informative source of social information.