Bit by Bit: Social Research in the Digital Age

Abstract

The implications and potential for social science arising from the availability of large amounts of data have been comparatively overlooked. This book provides a welcome insight into what can be achieved, and also an exemplar curriculum for graduate studies. Research with ‘big data’ has so far been dominated by computational capacities: the systems to process, analyse and visualize large data sets and the human skills to code, collate and interpret. The author is particularly well positioned because of his experience working in academia, government and the private sector. Salganik's text is structured by four broad approaches to digital data collection: experiments, observation, surveys and mass collaboration, each based round practical discussions of published work. He explains the challenges for each that are raised by the digital setting and new approaches, and concludes with a discussion of emerging ethical problems. The book is novel in drawing out the principles and practicalities as a guide for researchers. Conceptualization and validation of on-line data are addressed directly, though practical cases showing good practice remain rather scarce. The examples highlight the weaknesses and hype of published analyses. These can be corrected when they represent flawed interpretation, but this is often much more difficult when the fault lies in the design. Salganik provides several clear proposals for the potential use of big data by social scientists and presents compelling arguments for particular alternative practices. One caveat is that his recommendation to teach computational and social scientists in a mixed group demands coding aptitude and a critical attitude from all. Digital research needs to be viewed as more than just doing what you did before but by using a computer, with the computational aspects as the only challenge. Cheaper data collection comes generally at a cost in quality, although that can be addressed by greater repetition and redundancy. Differences in population coverage, which can be socially skewed or geographically dispersed, represent new concerns. Digital surveys offer opportunities to enhance standard approaches by recognizing size and quality as complementary, and consideration of experimental power leads to different solutions such as more-complex nested designs. The long evolution of the book included an innovative on-line, open review process that has paid dividends above the usual publisher's peer review. A diverse group of readers made annotations on the hypothes.is platform, adding judgements, suggestions and corrections or marking confusions. Although the mass collaboration process might be particularly suited to the topic of digital social research, this whole platform is now openly available for wider use at www.openreviewtoolkit.org. References are contemporary and sufficiently compendious that some will be unfamiliar, so building a comprehensive background for each recent example. The only exception is the chapter on ethics, whose US focus omits international developments that are specific to other countries, yet still hints at the wider principles such as consent and context. Every chapter, even that on ethics, ends with exercises that have clearly been developed through practical teaching. All the exercises usefully indicate the challenges that are involved, whether mathematics, coding or data collection. Formal statistical work is restricted to chapter appendices but is detailed when that forms part of the substantive challenge and is commendably clear and insightful. Consideration of sparseness in social preferences is enlightening, and the introduction to experimental design is exemplary. Topics like the separation of sampling from coverage are widely applicable in other settings, making the book a good choice for a general course on social research. The informed lack of hype regarding ‘big data’ enables critical consideration of the strengths and weaknesses of new opportunities for social research in the private sector as well as academia. Undergraduates pursuing extended research projects may find chapters useful but every dimension is advanced—ethics, coding and causal reasoning. This is, overall, an excellent book for graduate students and professionals needing to understand the implications of the emerging digital context for social research.