Twitter has announced a call for research submissions that helps them “identify indicators of conversational health that are even more specific to Twitter and its impact.” I expressed my skepticism about this on Facebook and so I am writing up some notes here.
‘Conversations’ are one way to examine interactions on social media. We looked at conversations as the unit of analysis in our Turnbull paper due out in MIA I think in a few months (based on our ANZCA paper). A simple point to make is that the ‘public conversation’ is not the same thing as ‘conversations on Twitter’ or even ‘Twitter publics’. Sure, there are conversations that happen entirely on Twitter (say a Trump tweet and reaction and cross talk), but these are not very useful as the basis of assessing public conversations. How Twitter users produce openings on other spaces, so that the circulation of discourse is necessarily cross-platform. The philosophy behind Cortico’s general approach looks interesting, but Twitter and Cortico will need to partner with other platforms.
There are broadly two ways to map the circulation of discourse. The first is derived from Bourdieu and maps a ‘field’ based on the social interactions between actors and the analytical construction of what is valued in the field (doxa). I think this is inherently flawed because of the reliance on a notion of faith (as in good or bad faith). Bourdieu’s Manet lectures are clear on this. The second is derived from Foucault and maps the discursive regularities between statements and the analytical construction is regarding the conditions of possibility based on ‘authority’ and composition of power relations (dispositif). What Foucault broadly called ‘eventalization’ (only ever in interviews, so the method has to be reverse engineered across a range of works). Interestingly, network graphing techniques seem to be aligned with ‘eventalization’ until you realise that they mostly rely on the providence of digital objects and platform-based network relations between them. There have been few attempts to map networks of discourse in spite of the platforms as this multiplies the work exponentially.
Analysing discourse in terms of the ‘health’ of conversations assumes a normative dimension that I think smuggles in assumptions about the good faith of actors. There are two problems here. First, analytics are unlikely to indicate how a particular user is ‘blinkered’, and therefore has an extremely constrained degree of freedom (in the systems theory sense), what Guattari called a low co-efficient of transversality or Warner might talk about in terms of the character of reflexivity. They will instead show how such blinkered users belong to tribes, because of the discursive coherency and affective congruence of discourse. So what? Second, Twitter does not appear to want to operate upon the good or bad faith of actors, and therefore take obvious steps to reduce the weaponised use of the platform (such as reduced functionality for new accounts until thresholds of participation are passed, such as number of followers or interactions). Getting over the normative assumptions about the good faith of users is an important first step.