My colleagues at the Digital Methods Initiative (Erik Borra, Natalia Sanchez-Querubin and Sophie Waterloo) and I just submitted an abstract for a social media theory and methods conference featuring great names in this space: Jean Burgess (Queensland University of Technology), Axel Bruns (Queensland University of Technology), Greg Elmer (Ryerson University) and Ganaele Langlois (U. of Ontario Institute of Technology).
The paper is called “Amazon as a Research Engine: Best Selling Issues in the Climate Change Debate” and proposes a protocol for repuposing Amazon.com as tool for debate mapping. Below is our abstract:
In this paper it is investigated how Amazon.com, the world’s largest online retailer and one of the leading product search engines, can be used for social and cultural analysis. By way of an info-political critique, Rogers (2013) tracks the state of issues or debates on the web by examining how Google authors stories through its search engine results. It is this move, going from search to research, that we take as a point of departure in our paper and that we use in order to work towards repuposing Amazon.com as tool for debate mapping.
We make use of Amazon’s “Best Sellers Rank” to compile lists of best selling books and their metadata, and repurpose them to study issue formatting and uptake in a debate. Formatting here refers both to how a larger topic is organised into sub-issues (i.e. types of language and how they are associated with other more tangible forms such as actors or places), and how the built-in analytics of online devices pre-order data which can be used for social research (Marres and Weltevrede 2013). Uptake refers to the shifting popularity of these specific formats or sub-issues in a debate over time.
Other studies have used the Amazon recommendation and ranking system to identify and profile reader groups and their commitments in particular. For example, Krebs (1999, 2003) uses the concepts of “communities of interests” and “buddy books” to describe these networks. Amazon itself attempted to characterize political tendencies in the United States, based on the purchases of political books state by state (2012). User purchases were employed as a “social metric” of agreement or interest. In this paper we repurpose the Amazon Best Sellers Rank not to profile the interests of readers but the uptake of issues instead, and describe the state of a debate according to its most popular and relevant formats. Identifying best selling books is used here as a descriptive strategy to map interest towards specific areas in a large debate in a determined moment in time.
Our study focuses on the debate taking place around climate change. The United Nations Framework Convention on Climate Change identifies two responses to climate change: mitigation by reducing greenhouse-gas emissions and adaptation to its impacts (Klein 2007). At the same time, the dissonant discourse of skepticism about climate change has acquired somewhat of a mainstream position in academic literature and the media (Niederer 2013). We take these three main approaches to climate change and examine how they are formatted and taken up in mainstream literature on Amazon.com. We ask: which are the best selling issues in the climate change debate and its approaches? What formats are more in demand on the market? Do the sub-issues differ for each approach to climate change? How does the uptake of these issues evolve over time?
These questions are operationalized through three methods. Firstly, we approach Amazon as a collection-making device. We developed a tool, the “Amazon’s format finder,” which allows users to query Amazon for particular keywords via the Product Advertising API and to retrieve for each query a list of the top 100 books titles and associated metadata, ranked according to their Best Sellers Rank. Similarly to headline analysis and scientometrics, we analyze books through their titles, and groups of titles through their components (keywords) and search for patterns amongst them (repetition, overlap, relations). As titles synthesize the main ideas and orientation of a text, and signal the importance of the particular topics, they are an appropriate object to study formating. The inclusion of the Best Sellers Rank adds, to what would otherwise be an analysis of media production, the element of consumption, and locates the formating of a debate as taking place in within a ‘marketplace for issues’.
We then proceed to map the best selling issues for each approach by querying “scepticism”, “adaptation” and “mitigation”, and capturing the keywords present in the titles of the 300 resulting books. The keywords are then clustered around themes. For example, the theme “water” is present in both books returned for the query “mitigation” and for “adaptation’, although the theme is paired –and therefore formatted– with the keywords “crisis” in the case of mitigation and with “governance” in the case of adaptation. Our visualizations accompanying the analysis shows where the three approaches overlap, compete and meet in terms of best selling issues (for an example, see annex I).
Secondly, the captured keywords are associated with the year of publication of the book they were extracted from. The result is an timeline, which indicates the freshness of best selling issues. Best selling scepticism issues today are included in both older and newer works, while popular adaptation and mitigation issues seem to be focused on more recently published works.
Thirdly, we use Amazon’s query autocomplete system as another way to study keyword uptake: which sub-issues are popular enough to be included in Amazon’s autocomplete list? The term “climate change” is inserted in Amazon’s search box and all subsequent suggested queries in alphabetical order are captured. The analysis of these queries shows “adaptation” and several keywords associated with “skepticism” (such as “hoax” and “denial”) are suggested from the outset. “Mitigation” is absent from this list, perhaps indicating a less popular or effective terminology or the decrease of interest in this particular sub-issue.
Just as the top results in Google have an info-political value, the best selling books in Amazon likewise allow us to glean insights of what issues drive the debate and about the biases in their representation. The absence in our mapping of issues that are relevant but perhaps not marketable, such as human rights or the victims of climate change, is telling as well. By exploring the uptake of these issues over time it is possible to see the shifts in the debate as well as the volatility of specific issues. As Amazon’s API targets different countries, we are able to do cross-country comparisons, localizing the debate as well. Our method thus allows for a critical inquiry into the influence of a recommendation engine on the uptake of specific formats in popular debates across countries and moments of time.
Flood, A., 2012. Republican landslide in Amazon book vote. the Guardian. Available at: http://www.theguardian.com/books/2012/aug/22/republican-landslide-us-book-vote [Accessed July 30, 2013].
Klein, R.J. et al., 2007. Inter-relationships between adaptation and mitigation. Available at: http://su.diva-portal.org/smash/record.jsf?pid=diva2:180448 [Accessed July 30, 2013].
Krebs, V., 1999. The Social Life of Books: Visualizing Communities of Interest via Purchase Patterns on the WWW. Available at http://www.orgnet.com/booknet.html [Accessed July 30, 2013].
Krebs, V., 2003. Proxy Networks –Analyzing One Network To Reveal Another. Bulletin de Méthodologie Sociologique, 79(1), pp.61–70.
Marres, N. & Weltevrede, E., 2013. SCRAPING THE SOCIAL? Issues in live social research. Journal of Cultural Economy, (ahead-of-print), pp.1–23.
Niederer, S., “Global warming is not a crisis!”: Studying climate change skepticism on the Web. Available at: http://www.necsus-ejms.org/global-warming-is-not-a-crisis-studying-climate-change-skepticism-on-the-web/ [Accessed July 30, 2013].
Rogers, R., 2013. Digital Methods. The MIT Press: Cambridge.
Van Couvering, E., 2007. Is relevance relevant? Market, science, and war: Discourses of search engine quality. Journal of Computer-Mediated Communication, 12(3), pp.866–887.