The internet is becoming increasingly visual, but social computing research and methodological training has relied heavily on textual methods. Methodological innovation is needed to study visual social data, including problematic information (mis- and disinformation, propaganda, hate, AI slop, etc). Contending with this, we present a framework for conducting grounded, interpretive, computationally supported, mixed-method research on collections of visual social media data. We developed this framework while grappling with the ethical, logistical, and methodological challenges of conducting in-depth analysis of potentially harmful visual content while caring for our research team. We document our framework components of visual grammars, human analysis, and computationally supported analysis with an umbrella commitment to care and its use in three empirical case studies. We also provide recommendations and implications for the HCI community in embracing training in and the advancing of visual methods and research, including a sensitizing concept of visual integrity.

Meta’s brief release of its social AI chatbots highlighted the challenges of studying systems that are both short-lived and relationally complex. In response, we conducted a 10-day collaborative autoethnography to rapidly and meaningfully engage with the product. As the first application of this method to social AI chatbots, our study demonstrates its value for examining ephemeral and emotionally complex AI systems.

Social media platforms are known to facilitate sudden bursts of attention on individual pieces of content and their creators — dynamics often referred to as ''going viral''. However, questions remain about how these moments impact the individuals who experience them, especially within algorithmically-mediated, video-sharing platforms like TikTok. In this study, we seek to better understand the experiences of creators who received a burst of attention, as identified by their participation in a trend highlighting an earlier video that had ''blown up.'' Through mixed-methods analysis of TikTok trace data and interviews, we show that massive surges in attention can have significant, varied, and long-lasting impacts on creators — within the platform and beyond. These include a short-term increase in views and sharing more similar content, and posting more content responding to other content on the platform. We also show how algorithmically-mediated bursts — or ''boosts'' — of attention led to shifts in creators' conceptualizations of themselves, the platform, and their audiences. Our work contributes empirical and conceptual insights into online fame and celebrity, how audiences and influencers interact, and how the affordances and cultures of TikTok shape how sudden attention is experienced on the platform.

As a research tradition, participatory design (PD) tends to focus on power dynamics where researchers hold greater power than participants. This paper uses design fiction to consider what this tendency overlooks by examining settings where participants may exist in multiple power relationships simultaneously implicated by the research, specifically focusing on the contexts of misinformation, disinformation, and online hate (M/D/OH). Drawing from existing literature in M/D/OH, we present a series of imaginary method abstracts that prompt questions for researchers to reflect on as they adapt PD techniques for new, different contexts. We highlight three value tensions—authenticity, reciprocity, and impact—integral to sustaining a concern for responsibility in PD scholarship. We end with reflections and potential considerations for responsibly applying PD and design fiction methods in M/D/OH settings.

The 2020 United States (US) presidential election was - and has continued to be - the focus of pervasive and persistent mis- and disinformation spreading through our media ecosystems, including social media. This event has driven the collection and analysis of large, directed social network datasets, but such datasets can resist intuitive understanding. In such large datasets, the overwhelming number of nodes and edges present in typical representations create visual artifacts, such as densely overlapping edges and tightly-packed formations of low-degree nodes, which obscure many features of more practical interest. We apply a method, coengagement transformations, to convert such networks of social data into tractable images. Intuitively, this approach allows for parameterized network visualizations that make shared audiences of engaged viewers salient to viewers. Using the interpretative capabilities of this method, we perform an extensive case study of the 2020 United States presidential election on Twitter, contributing an empirical analysis of coengagement. By creating and contrasting different networks at different parameter sets, we define and characterize several structures in this discourse network, including bridging accounts, satellite audiences, and followback communities. We discuss the importance and implications of these empirical network features in this context. In addition, we release open-source code for creating coengagement networks from Twitter and other structured interaction data.