In recent years, large-scale analysis of log data from digital devices - often termed ""big data analysis"" (Lazer, Kennedy, King, & Vespignani, 2014) - have taken hold in the field of internet research. Through Application Programming Interfaces (APIs) and commercial measurement, scholars have been able to analyze social media users (Freelon 2014) and web audiences (Taneja, 2016) on an uprecedented scale. And by developing digital research tools, scholars have been able to track individuals across websites (Menchen-Trevino, 2013) and mobile applications (Ørmen & Thorhauge 2015) in greater detail than ever before. Big data analysis holds unique potential for studying communication in depth and across many individuals (see e.g. Boase & Ling, 2013; Prior, 2013). At the same time, this approach introduces new methodological challenges in the transparency of data collection (Webster, 2014), sampling of participants and validity of conclusions (Rieder, Abdulla, Poell, Woltering, & Zack, 2015). Firstly, data aggregation is typically designed for commercial rather than academic purposes. The type of data included as well as how it is presented depend in large part on the business interests of measurement and advertisement companies (Webster, 2014). Secondly, when relying on this kind of secondary data it can be difficult to validate the output or techniques used to generate the data (Rieder, Abdulla, Poell, Woltering, & Zack, 2015). Thirdly, often the unit of analysis is media-centric, taking specific websites or social network pages as the empirical basis instead of individual users (Taneja, 2016). This makes it hard to untangle the behavior of real-world users from the aggregate trends. Lastly, variations in what users do might be so large that it is necessary to move from the aggregate to smaller groups of users to make meaningful inferences (Welles, 2014). Internet research is thus faced with a new research approach in big data analysis with potentials and perils that need to be discussed in combination with traditional approaches. This panel explores the role of big data analysis in relation to the wider repertoire of methods in internet research. The panel comprises four presentations that each sheds light on the complementarity of big data analysis with more traditional qualitative and quantitative methods. The first presentation opens the discussion with an overview of strategies for combining digital traces and commercial audience data with qualitative interviews and quantitative survey methods. The next presentation explores the potential of trace data to improve upon the experimental method. Researcher-collected data enables scholars to operate in a real-world setting, in contrast to a research lab, while obtaining informed consent from participants. The third presentation argues that large-scale audience data provide a unique perspective on internet use. By integrating census-level information about users with detailed traces of their behavior across websites, commercial audience data combines the strength of surveys and digital trace data respectively. Lastly, the fourth presentation shows how multi-institutional collaboration makes it possible do document social media activity (on Twitter) for a whole country (Australia) in a comprehensive manner. A feat not possible through other methods on a similar scale. Through these four presentations, the panel aims to situate big data analysis in the broader repertoire of internet research methods.
A DIVISION OF LABOR: THE ROLE OF BIG DATA ANALYSIS IN THE REPERTOIRE OF INTERNET RESEARCH METHODS