Developing Data Fluency Methods and Tools

Typically, most of the quantitative methods used by social media companies and computational social science researchers to understand the spread of mis- or disinformation online employ the following approach: 1) collect enormous amounts of data either via active user surveillance or by scraping public data on one platform, using technical tools and Application Programming Interfaces (APIs) that are not widely accessible or user-friendly; 2) track the spread of mis- and disinformation across a group of users and analyze the network of users or data for connections and clusters; 3) visualize this data in network maps; and 4) make claims about polarization and the effects of mis- and disinformation based on these visualizations. These techniques, although important to mapping the scale of mis- and disinformation, leave many questions unanswered regarding the why, how, and what for–all of which are better revealed through ethnographic studies, historical analyses, and close readings of content. As well, conventional techniques leave untouched the question of how opaque algorithmic recommendations affect the spread of mis- and disinformation by clustering users into what our PI has called “agitated clusters of comforting rage.”^[i]

Our approach brings together the strengths of computational, qualitative, and speculative methods and theories. Our suite of open source and user-friendly tools and infrastructures will enable researchers and community partners to collate, curate, analyze, and express archived data (English-language text and memes from social media platforms such as Twitter, Reddit, and Instagram) through APIs. These tools (see the Technologies to be Developed section and Appendix J) will be embedded within speculative and qualitative methods designed to offer more complex and nuanced understandings of how and why discriminatory, false and misleading information spreads. For example, we will extend approaches to network analysis by incorporating research persona and other speculative walkthrough methods (see Appendix E) that deploy fictional characters on social media platforms to explore disinformation processes. We will also couple Imageflow, the tool we have developed to discover meme clusters, with arts-based workshops. We will make these methods publicly available through the DDI web app and feature these analyses in our Data Fluencies Exhibition (see below). Further, we will work with other humanities researchers to incorporate, expand on, and supplement these methods through research development workshops, small grants, and curriculum development (see stream 4).

[i] Chun, Discriminating Data, 27.