Methods for big data in social sciences

Abstract

The diffusion of digital technologies and social networks has multiplied the forms of digital data that can be employed for social research. The main two forms are native digital data, which are produced in social networks, search engines, or blogging, and digitized data, which are analog data transformed into digital (Rogers, 2013). Big data are originally produced in the Internet. They allow for analyzing behaviors without interfering with individuals (Webb et al., 1966). An example is the data used in web platforms analytics, such as Google Correlate, whose purpose is to reveal the co-occurrences associated with a keyword searched through the Google search engine. This tool helped to predict the flu epidemic in the US, well before the US Centre for Disease Control and Prevention (Ginsberg et al., 2009). This example demonstrates that digital web platforms enable innovations in data analysis. Another example of native digital data is the data voluntarily uploaded on social networks, blogs, and websites. These are mainly textual or visual (images and videos), often unstructured. A third example is transactional data and the Internet of things. Transactions made through digital devices, such as smart-phones, scanners, tablets, and cards with chips (credit cards, shopping cards) produce data with some structure. These data comprise metadata (date, time, duration, or expenditures) associated with transactions. The objects connected to the Internet (the Internet of things), such as sensors for health monitoring, house automation, and driving aid, usually produce structured data, which can be organized and analyzed. Digitized data previously existed in analog form, for example images, videos, and scanned or digitally photographed documents uploaded on the web, such as museum collections or libraries available on-line. Digital humanities have converted this material into digital form. Another example is the surveys assisted by computers, where the data are inserted into digital databases. Web surveys now are conducted through the Internet (by e-mail) (Amaturo and Aragona, 2016), and allow for reaching a large sample with a small budget.

Editorial: Big data and machine learning in sociology

“…Opportunities to collect and use data of previously unknown mass, granularity, and complexity, in new formats and based on non-scientific and unknown data-generating processes require analytical models that adequately address these data characteristics (e.g., Amaturo and Aragona, 2019 ; Edelmann et al, 2020 ). In recent years, impressive computer hardware innovations regarding storage capacities, computing power, interconnectedness, task division, and data transmission evoked the development of such computationally intensive statistical software solutions, creating an algorithmic culture of statistical modeling without assuming an underlying stochastic data model as in the traditional statistical modeling culture (Breiman, 2001 ).…”