With the increasing digitization of personal data (Schüll 2016), behavioral data of interpersonal interactions are available at increasingly precise granularity and large scale. Granularity refers to these types of data having the capacity to encode data at observational intervals (e.g., second-by-second) that are unattainable for other research methods, while the scale of such data can often be multiple orders of magnitude larger than would be possible in interview or survey research. For example, using wearable sensors that recorded the “close proximity interactions” of students and staff in a school throughout the day, scholars were able to generate a highly granular map of behavioral interactions for the entire school (Salathé et al. 2010), which was then used to develop a model for improving intervention efficacy in a flu outbreak (Salathé and Jones 2010).
In a similar vein, but less obtrusive to deploy, app-based recording of interpersonal interactions has been developed for smart-phones (Eagle, Pentland, and Lazer 2009). These studies provide highly granular data of participants’ interactions with others in the form of phone calls, text messages, or even geo-location based estimates of their face-to-face interactions. These interactions can then be used to estimate social relationships (Raeder et al. 2011), which in-turn have been examined to show, for example both social influence and homophilous selection on drinking behavior among a cohort of undergraduates (Wang, Hachen, and Lizardo 2013). With the cooperation of cell phone providers, similar analyses can be conducted at the scale of entire countries (Wesolowski et al. 2012).
It is important to consider how readily such behavioral trace data compare to more traditional sociocentric surveys (adams 2010). Mastrandrea (2015) compare the resulting data from behavioral trace sources, sociometric surveys and social interaction diaries. They find dyadic differences in what gets recorded across platforms, (e.g., shorter duration interactions are under-represented in the diaries and surveys), and that a variable relationship between contact patterns and self-reported relationships (e.g., reported friendships correspond with both short and long duration contacts). Notably however, several constant structural features of the network (e.g., homophily estimates) were relatively consistent across the modalities (Mastrandrea 2015). Undoubtedly, similar comparisons will continue to be important as new platforms (such as Network Canvas) become increasingly common research tools.