Getting beyond WEIRD with Big Data
Social science has a weird problem, or the WEIRD problem, as Joseph Henrich and his colleagues called it in their 2010 critique of sampling biases in behavioral and social science research. They note that upwards of 90% of behavioral science research is conducted using participants from Western, Educated, Industrialized, Rich and Democratic (WEIRD) societies, and that undergraduate students attending American universities are 4,000 times more likely to be sampled into social science research than the average person. This presents some obvious challenges for the external validity of social science results. You would be hard pressed to find a scientist who would hold steadfast the idea that we should really be basing our fundamental understandings about human behavior on the performance of college undergraduates. But, any social scientist – especially an early career social scientist trying to establish a research program on a limited budget and under great time pressure – will tell you that sometimes, external validity is the price of efficiency. To get anything done at all, we have to use the resources that are most readily available to us. And the one resource university faculty have in abundance is a pool of willing, often free, undergraduate research participants.
Big Data are poised to change the resource equation. Although far from comprehensive, most Big Data datasets are less WEIRD than the college undergraduate samples that many social scientists have come to rely on. Already, researchers are making great use of Big Data to remedy the WEIRD problem – for example, by conducting cross-cultural research comparing Eastern and Western online communities, or following the twitter habits of citizens living in authoritarian states. In my article in the Early Career Researcher Forum, I encourage social scientists to seek out Big Data generated by participants who have been historically overlooked, underrepresented or excluded from social science research, even if that means actively whittling down a Big Data dataset until it is quite small. By choosing to make Big Data small, we can make science a little less WEIRD, and a little more inclusive, moving forward.
About the author
Brooke Foucault Welles is an Assistant Professor in the Department of Communication Studies and on the faculty of Network Science at Northeastern University. Her research focuses on networked communication, with particular emphasis on how people use online communication networks to facilitate personal and organizational goals. You can learn more about Brooke’s research on her website or on Twitter.