Goodreads Versus Amazon: The Effect of Decoupling Book Reviewing And Book Selling

Book reviewing is a commonplace activity on many ecommerce sites. However, it is nested within the broader context of book buying and selling. Goodreads, an online platform for social curation of book collections, provides an opportunity to observe on-line book reviewing in an environment that is not (at least overtly) focused on commercialization. In this study, we perform a careful comparative study of reviewer behavior and engagement in Goodreads and Amazon.com, constrained to a single genre (biography), including 21,394 books and 2.5 million reviews. We discover marked differences between the platforms that suggest disparate population composition and objectives of review-writing across the two platforms. Our findings suggest an important and generalizable principle: that two platforms engaging users on the same task (e.g., book review writing) may elicit quite different behavior depending on the implicit or explicit context and motivation present. User-contributed product and service reviews have become exceedingly common on a wide range of on-line platforms. This is due, in large part, to the significant value that such reviews carry for various stakeholders. For visitors to a site, reviews provide valuable information (typically in the form of recommendations); for platform owners, reviews drive traffic to the site; for review authors, writing reviews can feel altruistic and can also confer status within the online platform; and for producers of the products and services being reviewed, such reviews can influence sales — both on the platform and more broadly. Because on-line reviewing has become widespread and carries both economic and social value, significant research has been done to characterize and understand reviewing behavior, with a great deal of work focusing on topics such as review authenticity (Ott, Cardie, and Hancock 2012; et al. 2013), review relevance estimation (Martin and Pu 2014), and rating prediction (McAuley and Leskovec 2013). Of course, insofar as reviews are intended to represent the candid views of users (who have experienced the product/service in some way), a central question in this domain is the extent to which reviews of the same products and product class are strongly conditioned by the context in Copyright c © 2015, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved. which they are written. Given two sites eliciting reviews on the same products how similar or different might we expect their respective reviews to be? To our knowledge, very few studies have approached this question the most relevant work of which we are aware considered the effect of exogenous factors (weather and demographics) on the content and style of recommendations (Bakhshi, Kanuparty, and Gilbert 2014). While relevant, prior work such as this has not addressed the question of how platform context shapes review writers and reviews themselves. In this study, we investigate this question within the context of book reviews. Books continue to represent one of the largest sectors of the culture industry, with most recent reported annual U.S. sales of just over 27 billion USD (AAP 2013); U.S. music sales by comparison were just under 4.5 billion USD (IFPI 2014). Amazon and Goodreads are the two largest English-language platforms that allow users to write and publicly share reviews of books. The stated purpose of these sites, however, are quite different. Amazon is an e-commerce site, such that book reviews presumably inform the book purchases that web visitors make. Goodreads, by contrast, is a book “social cataloging” site in which users can keep track of books they have read/are reading/would like to read, organize books into collections, and add comments to books (which function as reviews insofar as comments are public to the entire Goodreads community, include star ratings, and are aggregated at the book-level). Thus, a key difference between Goodreads’ and Amazon’s e-commerce site is that Goodreads does not link the review-writing process to purchasing decisions in any overt way (the interested Goodreads user can find a link to booksellers, but this information is very discretely positioned on the book’s webpage). It is worth noting that while Goodreads is owned by Amazon, there appears to be little or no integration between the two services (other than a more prominent link to Amazon.com than to other booksellers). By studying Goodreads and Amazon reviewing behavior at large-scale, we have the opportunity to observe how overt review-commercialization impacts review writing and review construction. Notably, while work has been done on Amazon reviews, this is the first published study to perform large-scale analysis of any kind on Goodreads data (e.g., (Chevalier and Mayzlin 2006; Hu, Jok, and Reddy 2014)). In this first exploratory study, we limit our analysis to a Proceedings of the Ninth International AAAI Conference on Web and Social Media