This post is a digital appendix for one of my chapters in the book, co-authored with Anne Galloway and Theresa Sauter. The chapter is “Hashtag as hybrid forum: the case of #agchatoz” (PDF), and it builds directly on work I presented in Amsterdam last year and blogged about here. Here’s the first part of the introduction:
This chapter imports Michel Callon’s model of the ‘hybrid forum’ (Callon et al, 2009, p. 18) into social media research, arguing that certain kinds of hashtag publics can be mapped onto this model. It explores this idea of the hashtag as hybrid forum through the worked example of #agchatoz—a hashtag used as both ‘meetup’ organizer for Australian farmers and other stakeholders in Australian agriculture, and as a topic marker for general discussion of related issues. Applying the principles and techniques of digital methods (Rogers, 2013), we employ a standard suite of analytics to a longitudinal dataset of #agchatoz tweets. The results are used not only to describe various elements and dynamics of this hashtag, but also to experiment with the articulation of such approaches with the theoretical model of the hybrid forum, as well as exploring the ways that controversies animate and transform such forums as part of the emergence and cross-pollination of issue publics.
The digital appendix is a way of getting around some of the limitations of print formats for representing digital methods – in particular, the need to share data in some form, as well as to provide rich media data visualisations at sufficient resolution that they can be properly represented and interrogated as an integral part of the research, rather than being reduced to gestural or illustrative elements.
To that end, here I’m sharing some of the data tables and social network maps that are discussed in the chapter, each annotated with a short description and cross-referenced to the point in the text where it appears.
|Text Reference||Title and link to file||Description|
|Full text||Pre-press version of the chapter (PDF)|
|Figure 1||Tweets per day (tweets per day.png)||Number of tweets collected per day – a measure of activity over time and a way of identifying significant ‘peaks’ and ‘bursts’ in engagement|
|Figure 2||Tweets and unique accounts per day (tweets vs unique users per day.png)||Number of unique users per day in relation to number of tweets per day – a way of understanding which of the peaks remained confined to a small group of core accounts, and whether any peaks involved a significantly higher number of accounts; an indicator of topic diversity.|
|Footnote 4||List of most-shared URLs (Google spreadsheet)||Annotated list of the URLs that appear most frequently across the dataset (following manual resolution of top 50 working short URLs only). Useful as a preliminary indication of topics, organisations and media resources in play in an issue public.|
|Figure 3||Follower-followee network, nodes sized according to in-degree, or number of #agchatoz followers (fig3_indegree)||See notes on gephi social network visualisations below|
|Figure 4||Nodes sized according to out-degree, or number of #agchatoz
|See notes on gephi social network visualisations below|
|Figure 5||Nodes sized according to total overall Twitter followers (fig5_totalfollowers)||See notes on gephi social network visualisations below|
|Figure 6||Nodes sized according to number of #agchatoz tweets (fig6_agchatoztweets)||See notes on gephi social network visualisations below|
|Figure 7||Nodes sized according to number of all-time tweets (fig7_totalstatuscount)||See notes on gephi social network visualisations below|
|Footnote 8||List of hashtags that co-occurred with #agchatoz (Google spreadsheet)||Useful as an indication of intersecting issues and issue publics|
Notes on the #agchatoz dataset
We collected data using a custom installation of the YourTwapperkeeper tool (now superseded in our research by TCAT) and the keyword ‘#agchatoz’ (with the hash symbol included). The dataset underlying the analysis in this paper covers the period 19 July 2013–19 February 2014. It comprises 73,218 tweets. Because of Twitter’s strict restrictions on the public sharing of research datasets as well as ethical and privacy considerations, I am not sharing the underlying data set in its ‘raw‘ form.
Notes on the social network maps
The network graphs were produced by querying the Twitter API for the follower and friends lists for each username in the #agchatoz dataset as at April 2014. The resulting network contained 14,576 and 561,853 edges (based on follower-followee relationships). For each account, overall status counts, overall followers, and number of tweets in the #agchatoz dataset were added as supplementary node attributes. The network was spatialised in Gephi using the Force Atlas 2 algorithm. In order to focus the visualisation on persistent clusters of accounts (and removing accounts that participate in the hashtag without following each other), the network map was filtered so that only nodes with a degree of 7 (a combined total of 7 follower or followee connections) remained visible. The resulting map contains 6196 nodes (only 42.5% of the nodes in the original data set), but retains 559,827 (99.64%) of the original edges. Gephi’s modularity algorithm was used to calculate and then to colour ‘communities’ whose members have higher-than-random affinity, resulting in a total of six visible clusters and a modularity score of 3.9, leaving the tool at the default resolution setting of 1.0. Manual review of the profiles attached to the Twitter accounts that clustered together in these ‘communities’ was used to identify some common characteristics; lists of participant account IDs were subjected to an additional manual coding exercise to test the applicability of the resulting labels.