Frontiers of Science in Social Media Research: Computational and Experimental Methods to Study Digital Democracy

The electoral success of extremist politicians, the networked spread of disinformation, and an uptick in hate speech all relate to widespread social media use. Worryingly, the European political research community remains under equipped to study these phenomena, which challenge social cohesion and traditional models of state governance (Borucki & Schuenemann, 2019). Theoretically, scholars struggle to keep up with the rapidly changing social media landscape (Bossetta, 2018), and methodological difficulties in big data analysis limit researchers’ ability to produce robust and reproducible findings (Stieglitz et al., 2019). Compounding these challenges, social media platforms have recently imposed heavy restrictions on researchers’ access to data (Bruns, 2019; Walker et al., 2019; Freelon, 2018). Currently, three methodological approaches guide this growing field of research:  survey methods  computational methods  experimental methods. However, each approach has critical drawbacks that limit their contributions to science. While the primary method for studying social media is through surveys, research finds that citizens’ recall of self-reported behaviour does not align with their online activity (Vraga et al., 2016; Haenschen, 2019). Meanwhile, the computational analysis of digital data suffers from non-representative and incomplete datasets (Tromble, 2019), as well as the inconsistent application of research ethics (Shah et al., 2015; Tromble, 2021). Lastly, experimental designs that create ‘mock-ups’ of social media platforms fail to fully replicate these environments, because they cannot incorporate the personalized algorithms that these platforms offer (DeVito, 2016). Therefore, interdisciplinary and innovative solutions are needed to solve these methodological challenges and shed light on the digital aspects of modern democracy. Techniques borrowed from computer science, such as text mining, machine learning, and social network analysis open up new pathways for data collection and analysis. Inspiration can also be drawn from the behavioural sciences, where controlled experiments isolate how social media affects citizens’ cognitive processing and attitude formation. However, few scholarly arenas have focused on how computational methods and experiments can be combined to maximize the value of social media research and compensate for the limitations of each method. Knowing how to best synergize the findings from various methodological approaches will therefore aid scholars in their descriptive and prescriptive assessments of social media’s impact on democratic processes. This Workshop pushes forward an interdisciplinary methodological agenda to foster academic discussion at the intersection of political science, data science, and experimental methods. Paper proposals should involve either computational methods or experiments to address the overarching research questions: 1. How can the insights from computational social science inform experimental research, and vice versa? 2. What new methodologies, or innovative combination of existing methodologies, can be harnessed to maximize the findings of social media research in ways that promote our understanding of society, politics, and democracy?