Paper long abstract:

Since the mid-2000's, the abundance of digital traces of conversations has been viewed as a crucial opportunity by a fast-growing economic sector, essentially represented by software editors, which claims to access thanks to this data to massive and spontaneous opinions - as opposed to the low-scale and solicited opinions of the polling industry. These companies thus have to deal with the same issues than does the academic research; one of them, well identified by Rogers (2009), is the inscription of the opinions within social groups, given the lack of socio-demographic information about online individuals. Consequently, they must build alternative epistemologies to guarantee - in new ways - the validity of the knowledge they produce about online opinion.

In this communication we will describe how this issue is addressed by a French company named Linkfluence, by showing how, using the same mapping methods used in Digital Humanities, they reorder the online social space around artifacts called "communities", built on the aggregation of the websites' hypertext profiles, allowing them to inscribe opinions into stabilized social forms. Along with Desrosières (2008) and Didier (2009), we want to enter into the very procedures which generate categories of social organization. For this, we draw on theoretical papers produced by our actors and the literature they refer to, but also on a two-month observation conducted in the company's offices. By doing so, we aim to show how this practice of web mapping embeds a social order, based on fluid, elective and action-centered collectives named "communities".