ANZCA 2017

Dynamics of a Scandal: The Centrelink Robodebt Affair on Twitter

Axel Bruns, Brenda Moon, and Ehsan Dehghan

• 7 July 2017 – Australia New Zealand Communication Association conference, Sydney

Past months have seen considerable controversy over the operation of an algorithmically controlled debt recovery system for suspected overpayments by the Australian federal social services agency Centrelink. Incorrect debt recovery notices, often for considerable amounts, have been sent to a substantial number of Australian welfare recipients, while channels for the disputation and redress of such notices were often left unclear or, where available, were overloaded by the volume of calls from affected citizens.

Since the beginning of 2017, this matter has received considerable political, media, and public attention, ranging from heated parliamentary exchanges through investigative reporting on internal Centrelink case management practices to public debate in social media. The #notmydebt campaign, rallying around the eponymous Twitter hashtag but operating across a number of online platforms, has facilitated the collation of citizens’ experiences with Centrelink; in response, the targetted leaking by federal government sources of at least one vocal Centrelink critic’s personal details to the press has further inflamed the controversy.

This paper explores the dynamics of the ‘robodebt’ affair through the first half of 2017. We draw centrally on a major dataset of tweets relating to the affair that is drawn from the multi-institutional, collaborative TrISMA infrastructure (Bruns et al., 2016), which tracks all public tweets by the four million Australian Twitter users identified to date on a continuous basis; this provides the considerable advantage of being able to filter the full dataset of Australian tweets for posts referring to the issue without needing to track a pre-determined and immutable set of keywords only.

Instead, we assemble our dataset from those tweets matching a number of core terms (Centrelink, robodebt, #notmydebt, etc. – amounting to more than 300,000 tweets for January and February 2017 alone), identify further topical keywords from that set, and repeat that sampling process multiple times to establish a larger, more comprehensive set of posts. We also add to our dataset all those posts that were made in @reply conversations on Twitter immediately before or after our matching posts, even if these additional @replies did not themselves contain any matching keywords. This results in a more comprehensive picture of the full range and volume of conversations than would be possible to establish by other means.

We use this dataset to investigate the dynamics of the Centrelink affair: over time, we identify the major individual and institutional actors highlighted in the discussion; we examine the key external sources of new information being introduced (media reports, press statements, etc.) and assess their relative impact on continuing public debate; and we trace the changing popular framing of the affair by assessing the shift in descriptive language over time. Finally, we also correlate our observations with the detailed information about the structure of follower relations and thematic clusters in the Australian Twittersphere that is available from TrISMA, to examine which parts of the network are active at any one point of the debate.

References

Bruns, Axel, Jean Burgess, John Banks, Dian Tjondronegoro, Alex Dreiling, John Hartley, Tama Leaver, Anne Aly, Tim Highfield, Rowan Wilken, Ellie Rennie, Dean Lusher, Matthew Allen, David Marshall, Kristin Demetrious, and Troy Sadkowsky. TrISMA: Tracking Infrastructure for Social Media Analysis, 2016. http://trisma.org/.