How Editorial, Temporal and Social Biases Affect Online Food Popularity and Appreciation

Abstract

Measures of popularity and appreciation provide useful information for search and recommendation systems that facilitate access to growing amounts of user-generated content, such as online recipes. However, user rating and commenting behavior is not only influenced by the content itself, but also due to additional effects introduced by biases and contexts. Based on a large dataset of more than 400,000 online recipes, we investigate the nature of such biases and the impact on the number of ratings, comments and views. Our analysis shows that user feedback is significantly influenced by the recipe author's prior reputation, by friendship relations, similarity between user profiles, temporal and seasonal effects, and editorial choices. Furthermore, a regression analysis shows that for the number of ratings received by recipes in particular, an excellent fit can be obtained based on a combination of these biases. These results imply that the popularity of an item is heavily influenced by random bias introduced by various external factors that impact rating and commenting behavior in a relatively short time span after publication.