Between sex and data science, I can claim success in one much more than the other one. So let’s talk data science! Or to be precise; social data science.
First things first, it’s not social data science, it is social data science!
Do you see the difference between these? What I’m trying to say is that it’s Social data-science rather than social-data Science. It’s not data science applied to social data, it’s a data science that is social. And I don’t mean a data science that is talkative and fun to be around, I mean a data science that borrows from social sciences. Maybe it’d be better to call it social science data science or social data science science. How about computational social science?
Let me try again! We want to understand how human societies work - or how humans work in societies. Nothing has ever been more puzzling to humans than humans themselves and particularly the social aspects of human life. Throughout history, we have been fascinated by wars, religions, innovations, revolutions, cultures, literature, politics, economy, education - the list goes on. And the show never stops. When was the last time a human-phenomenon surprised you to a jaw-dropping level?
For me it was when the Instagram Egg received more than 53 million likes!
We want to understand how such things happen. And that’s why we have built social science departments across the globe and teach students about social theory, research methods, experimentation, survey studies, and statistics. However, though we can explain social phenomena, this mostly is many years after they’ve happened. In other words, we are not yet able to ‘predict’ social phenomena. Could you tell me which Instagram photo is the next one to go viral?
Why is it so different to the natural sciences?
Natural sciences, on the other hand, have been more successful in predicting natural phenomena - at least in certain areas. It’s still impossible to predict the next big earthquake (though slightly less difficult to forecast the next hurricane). But want to know when and where you can see the next eclipse? No problem! Want to know how to prevent a wound infection? Sure thing! Want to know how to cook meth? Let’s not get SAGE in trouble!
I’m not claiming that social problems are as easy to solve as natural problems, but I believe the massive difference we see in social and natural sciences in this respect is mostly due to two factors; data and methodology.
Natural sciences started to grow and experience success only when a massive paradigm shift happened around the 9th and 10th centuries. It was when Muslim scientists in the Middle East started to use experiments and quantification to test and distinguish between competing theories. This practice appears very natural to us today, but had been mostly seen as an abnormality in Medieval Europe - until the 12th century translation of Arabic text books into Latin and the work by English philosopher Roger Bacon facilitated the formation of what today we call the scientific methods.
Scientific methods are based on a repeating cycle of observation, hypothesis forming, prediction, and experimentation. How does this differ to the social sciences? The short answer to this question is that experimentation and quantification are much more complicated when it comes to human subjects.
Isolating different factors and repetition in an experiment in natural systems are much easier than in social systems. We cannot put groups of humans in cages to isolate them from any external factor to have a controlled experiment (the TV show Love Island being an exception!). Who wants to repeat the Second World War for the sake of a better statistics?
Another issue in social science research is privacy. In studying systems from afar, like stars and galaxies, precise quantitative observations are available as a tool to test different theories, even though scientists cannot have controlled experiments.
Big data and social data science – a change to observation
Privacy issues are still very important, but the way ‘observation’ can be conducted in understanding social systems has dramatically changed in recent years. This is due to the advent of large scale transactional data we produce on digital technologies - aka big data!
Social data science can be thought of as a way that a science involving social systems can be done in the 21st century. In modern times, there has not been any successful scientific endeavour which did not rely on data. Therefore, we needed to develop techniques that allow us analyse large-scale datasets to explore patterns and relationships - aka data science!
In the Research Design in Social Data Science online course, we walk you through the steps that need to be taken to design a social data science research project. We discuss the dos and don’ts of studying a social phenomenon based on large scale transactional data in an ethical framework. We provide an overview of the methodologies that are available to analyse the data, some adopted from qualitative social science methods, and how to draw conclusions from your research that are trustable and can be considered as an addition to the knowledge on human social behaviour.
 David C. Lindberg (1980), Science in the Middle Ages, University of Chicago Press, p. 21.
 Ackerman, James S. (1978), "Leonardo's Eye", Journal of the Warburg & Courtauld Institutes, Vol. 41