AANZCA 2024

Untangling the Furball: A Practice Mapping Approach to the Analysis of Multimodal Interactions in Social Networks

Axel Bruns, Kateryna Kasianenko, Vishnuprasad Padinjaredath Suresh, Ehsan Dehghan, and Laura Vodden

• 25 Nov. 2024 – AANZCA 2024 conference, Melbourne

Presentation Slides

Abstract

Social network analysis has become a key tool for the analysis of user actions and interactions on contemporary social media platforms, and beyond. Often, however, such analyses remain somewhat superficial, merely presenting the standard network graphs produced by the key visualisation algorithms implemented in popular network analysis tools like Gephi, or identifying the clusters of tightly connected accounts highlighted by popular modularity algorithms, without sufficient consideration of the limitations of such visualisation and analysis approaches. At worst, and especially in the hands of inexperienced researchers, such attempts to make sense of networks result in network ‘furballs’ of severely limited value; even if the network analysis and visualisation produces networks with more distinct features, however, they often fail to represent more than a handful of obvious patterns, reducing complex and multilayered action and interaction patterns into overly simplified network graphs and statistics.

Social network analysis – or, more to the point, the analysis of participant practices in social networks – need not remain constrained by these limitations. Rather, in this paper we present the concept of practice mapping, and the use of vector embeddings of network actions and interactions as a means to map such practices, as a framework for methodological advancement beyond the limitations of conventional network analysis and visualisation. In particular, as we demonstrate in this paper through a number of brief examples, the methodological framework we outline here has the potential to incorporate multiple distinct modes of interaction into a single practice map, independent of any one specific social media platform, and can also be further enriched with account-level attributes such as information gleaned from topic modelling, profile information, available demographic details, and other features. We argue that constructing this network map of practices can offer a view of the connections, ruptures, and overlaps between the doings and sayings of all, rather than only of the most active or influential members of a group of users. A map of practices, constructed in the way we outline, can also serve as a sampling frame for deeper contextual or interview analysis, adding the first-hand perspectives of practitioners to the birds-eye view offered by the map. In this paper we demonstrate the utility of this approach by using a dataset of Twitter debate about the Robodebt scandal from 2016 to 2023 as an example.