The analysis and visualization meeting was the last of the methodological meets in the SMR Convergence meeting series. In many ways, it was the most complex. First, analysis and visualization (even visual analytics) are distinct topics. But there is overlap. For example, it is common to use different visualizations during analysis and given the sheer volume of social media data, visual analytic tools can be important during analysis, as well as during other parts of the research lifecycle. Choices about analysis may be informed by visualizations and vice versa – both are key in communicating about a data set and what it means.
Another challenge of this meeting was how to get scholars from different disciplines to a common starting point rapidly in order to have a meaningful discussion about analysis of social media data. One cannot think about analysis without considering the research question, the available sample and how the data were collected, the construction, reliability and validity of social media derived measures, and the computational models used. Therefore, a large goal when planning this meeting was to synthesize vocabulary and bring everyone to that common starting point. We also recognized that each field of research has different analysis techniques and different levels of familiarity with visual analytics. Putting these two topics into the same meeting provided us with the opportunity to think about analysis and visual analytics/visualization in new, synergistic ways.
We refer you to our white paper to understand the core content of the meeting. But we want to highlight a few things we collectively recognized as a group.
First, most of us come from traditions where researchers select a scale of analysis and stick with it – whether it is individuals, groups, or populations. Very few of us work across these scales. Social media is unique because it lets us conduct analyses at all of these scales, allowing for a broadening of our research traditions. Second, social media is not well suited for traditional social science causal research questions. It is better designed for descriptive questions. Casual questions are more connected to information flow, e.g. why some misinformation spreads faster. Third, social media gives us access to a wide range of data, allowing for many different types of analyses. For example, we have access to longitudinal data, network data, and spatial data. We have access to well-structured data values, text data, image data, and/or video data. These different forms of data allow for analyses that range from more traditional rectangular analyses to network analyses to predictive analyses. And visual analysis can be part of any of these, but is not being used consistently across disciplines. This is a real missed opportunity. Finally, researchers cannot figure out everything on their own. Social media is complicated and can be overwhelming. As we embark on research involving social media, we need to share our insights and tools so everyone can learn from them, debate about them, and improve on them.
To help with this process, the white paper we created contains different checklists researchers can use to think about different issues that arise when conducting social media research. The white paper then shows the checklists in action for three case studies. For each case study, we overview possible analyses using social media data, create an analysis plan, and highlight the challenges associated with the analysis.
Writing this final white paper really helped pull together the main highlights across all five meetings. We want to thank everyone that participated in all five meetings. It was both educational and fun. Our project team will continue to develop tools, templates, and infrastructure that helps improve research using social media data.