One-month Fellowship at ZeMKI, University of Bremen
Glad to announce that I was selected as a Visiting Research Fellow at ZeMKI, University of Bremen. I will be working mostly on systematising the comparative research methods I have been developing the past few years. The project description was as follows:
As a ZeMKI Visiting Research Fellow I propose to work on a research framework that I have been developing for the past four years at the Programme for Comparative Media Law and Policy, at the University of Oxford, and for the VOX-Pol Network of Excellence. This research framework – developed initially to explore the communicative dynamics of social media hate speech and online conflict – combines approaches from digital ethnography and techniques in computational text mining and network analysis to trace the global and local entanglements of contemporary digital cross-media conversations and practices. It focuses on three inter-related levels of analysis:
- Practices of production: The first level of analysis draws from methods in digital ethnography (virtual ethnography, data-driven digital ethnography, offline-online multi-sited ethnography) to examine the situated practices of production through which new kinds of content is produced and shared across digital platforms;
- Media content: The second level of analysis draws from approaches in computational text mining (such as deep learning word embeddings and LDA topic modelling) to examine the different types of textual and ideological content that is produced and how this content changes as it is shared across different platforms and audiences segments;
- Audiences and social networks: The third level of analysis draws on approaches from network analysis (dynamic social network analysis, multi-level network analysis) to examine the network topologies, key actors and communities that facilitate the sharing of this content across cross-media platforms.
To briefly illustrate what this kind of research approach involves, during my VOX-Pol research I analysed the cross-media conversation and practics of three major public Facebook groups related to the refugee crisis in Finland (racist anti-refugee, in- between discussion group and anti-racist). The research consisted of three steps:
- Firstly, I started of with digital ethnography to identify what key issues, themes and research problematics emerged from a qualitative engagement with the different social media communities and conversations in them;
- Secondly, after identifying what these key issues were, I then downloaded close to a million posts and comments from these groups. Based on a select number of key terms and topics identified through the initial ethnographic engagement, I then carried out computational text mining on the content shared and discussed in these groups. This involved experimenting with emerging methods in unsupervised machine learning (such as deep learning word embeddings to look at the word associations across selected keywords) as well as LDA topic modelling (to identify and compare key topics discussed in the different groups).
- Thirdly, I looked at what the underlying social networks and communities in these groups that enables the sharing of different kinds of content. This also allowed me to sub-divide the audience segments by overlaying textual attributes over networks to heuristically explore the inter-action between semantic content and social networks.
Finally, I repeated this data exploration of such “topical networks” until I was happy with both the qualitative and quantitative insights that emerged and was able to form more empirically-grounded insights and arguments. This three-pronged approach – iteratively combining qualitative and quantiative insight – thus allowed me to gain both a granular perspective to debates on hate speech around the refugee crisis discussed in these groups, the communicative patterns behind them as well as identify broader patterns and trends at a scale usually unavailable with qualitative methods.
With this approach in mind, I propose to expand this work in two ways at ZeMKI. Firstly, I want to systematise and share the approach with colleagues and other researchers working on different domains to identify its promises, weak spots, and challenges. And secondly, I hope to link what has been more so far a “heuristic working practice” to more grounded philosophical and theoretical debates in media and cultural studies. I hope to especially ground some of the more exploratory digital ethnographic approaches to ongoing debates in digital sociology and especially around theories such as assemblage theory, network theory, as well as work done at ZeMKI on communicative figurations and media ensembles.