The rapid development of Internet applications and devices has greatly reduced the costs of coordinating and participating in many social and cultural activities. Over the last fifteen years or so, there has emerged, through both corporate or individual initiatives, numerous large collectives producing information available to all. Beyond the paradigmatic example of Wikipedia, online video platforms, blog networks, and consumer reviews sites have together built rich data resources, based on free contributions and organized by site administrators and algorithms. These web-based platforms gather heterogeneous contributions from users, that they reconfigure through operations of selection and aggregation, then sort and shape this information in order to make it meaningful to their audience. Several terms have been used to describe this mechanism: “collective intelligence” (Surowiecki, 2005), “the wealth of networks” (Benkler 2006), and “wikinomics” (Tapscott and Williams, 2005). The analyses of these authors highlight the ability of such forums to create greater value from scattered individual contributions. They emphasize the efficiency of algorithms and the coordination of technical systems that enable the aggregation of subjective and local contributions into a larger whole that is relevant for users. Overall, these systems and the mathematical formula that support them, whether simple or complex (based on rankings, averages, recommendations, etc.), are able to build valuable assets from myriad heterogeneous elements produced without a larger collective aim.
Online consumer reviews (OCRs) are a good illustration of this phenomenon. First popularized by Amazon in the late 1990s, they have since become ubiquitous on the web. They are typically comprised of a combination of a rating (often out of five, and symbolized by stars) and a written review. A product’s overall evaluation is summarized by the average rating and the first few lines of some reviews, which the user can freely navigate. OCRs are now present on a variety of sites, particularly those platforms that specialize in collecting opinions (TripAdvisor, Yelp, LaFourchette) and e-commerce sites. They cover a wide variety of goods and services, from hotels and restaurants to funeral homes, as well as books, vacuum cleaners, schools, and everything in between. By bringing together a unified representation of scattered consumer voices, the rating-and-consumer-review system has clearly formed a large part of our collective digital intelligence. Indeed, the creators of these sites themselves often invoke democratic legitimacy by presenting themselves as the voice of ordinary consumers. Like in democratic elections, every consumer is allowed to vote, and all opinions are presumed equal. For example, the CEO of TripAdvisor has stated: “Online travel reviews have hugely changed the way the travellers can plan their holidays – they add an independent view of where to go and stay giving another level of assurance that their hard earned travel Euro is spent wisely. […] That’s the positive power of Internet democracy in action” (Kaufer, 2011). A further claim to legitimacy is the strong consumer appetite for these services, as a majority of Internet-users say they use them regularly; this has translated into tangible effects in many markets. Indeed, several marketing science and econometric studies have demonstrated a significant impact of OCRs on economic activity in sectors such as hotels, restaurants, and cultural consumption.
While it has received a lot of media commentary, the practice of rating and reviewing has received very little empirical research. The few that exist, mainly in sociology and organization studies, are schematically divided into two categories. The first investigates the motivations of those who frequently contribute comprehensive reviews, emphasizing the importance of recognition, skill development, and gratification: according to these studies, OCRs appear primarily to be the work of semi-professional evaluators (Pinch and Kessler, 2011), somewhat leaving ordinary contributors on the margins. A second category insists instead on the heterogeneity of scattered, subjective contributions, stressing the decisive role played by algorithms in constructing meaningful assessments, overall scores, and rankings (Orlikowski and Scott, 2014). These analyses support the perspective of broader reflections on collective intelligence, highlighting the crucial role of algorithmic synthesis, and calculations more generally, in the aggregation of subjectivites; they suggest that contributors are largely isolated, guided by an irreducible subjectivity, and, statistically speaking, independent.
Recently, web-based platforms such as OCR websites have held the attention of scholars for their capacity to organize the information and to make sense of the users’ contributions. By aggregating and sorting contributions through proprietary and often undisclosed algorithms, these websites have a strong power in shaping culture (Striphas, 2015). Through their algorithms, they are in position to redistribute valuations and preferences in many cultural and information industries, in a way that cannot be democratically discussed and disputed (Gillepsie, 2010; Morris, 2015). Though these analyses raise a crucial point – our ability to discuss what’s valuable in our cultures – they tend to presume that the effect of the algorithm is complete and undisputed. In a Foucaldian perspective, they stress the power of the web platforms to organize user’s information, and consider the algorithm as the result of an explicit strategy; conversely, users are mainly seen as passive subjects of this strategy. In this chapter, we try to qualify this perspective by underlining the role of users in the shaping of algorithmic valuation. As stated by Oudshoorn and Pinch (2003), “users matter” in the shaping of technology, and their actions shape the platforms in at least two ways. First, they interpret the information provided by the platforms, select and weigh it in a way that is not completely scripted by the site. These interpretations are based on their experience, and they have good reasons to adhere or not to the site’s valuations. Secondly, users shape the platform through their contributions, by following or not its guidelines, and by anticipating the effect of their actions. As a consequence, the “algorithmic” valuation is co-produced by the site and its users through a relationship that cannot be subsumed through pure alienation. Following MacKenzie (2011, 2014), the set of interpreting schemes and practices developed by users around the website can be called an “evaluation culture”.
In this chapter, we follow this user-centered perspective, by highlighting the collective practices and reflexivity of ordinary contributors. We show that the authors of such opinions do not give free rein to their subjectivity, but write in consideration of a specific target audience and/or website in mind. There exist common assumptions and norms concerning the proper format and content of an opinion, as well as standards governing what makes a contributor helpful, as well as a rating relevant. All of these standards can be described as part of the evaluation culture described by MacKenzie: the development of a new assessment tool is necessarily accompanied by the emergence of more or less coherent methods of interpretation, reading practices, and the manipulation of instruments. Rather than contributors primarily seeking recognition or consumers governed by their subjectivity, it is the figure of a common user who is reflexive, knowledgeable and accustomed to these services that we want to highlight here. In order to do this, we rely on a survey of contributors to the restaurant rating website LaFourchette (www.lafourchette.com), supplemented by contextual data from the web, as well as a survey of a representative sample of consumers.
 Beuscart J.S., Mellet K., «Shaping Consumers’ Online Voices: Algorithmic Apparatus or Evaluation Culture ?» , in Roberge J., Seyfert R., (ed.), Algorithmic Cultures: Essays on Meaning, Performance and New Technologies, London, Routledge. [presentation / pre-print version]