Len Metson

Society’s digital revolution has sparked an array of new questions for political scientists to answer. Is social media helpful or harmful to democracy? How can we mitigate online disinformation? What do dictatorships censor online? These questions are created by the explosion of data in the social world, and therefore data science is essential for political scientists to answer them. 

However, data science offers a second opportunity: it gives us an ability to answer old questions in new ways. Measurement is essential for political science. In order to test theories, scientists need to collect data through measuring concepts in the real world. The renowned American political scientist, Philip E. Converse, opens his classic chapter The Nature of Belief Systems in Mass Publics with a discussion of the problem of measurement. He challenges the belief that “what is important to study cannot be measured and that what can be measured is not important to study”. With characteristic rhetorical flair, he writes “no intellectual position is likely to become obsolete quite so rapidly as one that takes the current empirical capability as the limit of the possible”. Converse was writing about the opportunity to measure citizens' political ideologies through surveys, a method now synonymous with politics research. 

I believe techniques borrowed from data science have a similar revolutionary potential. Data scientists have become very good at quantifying typically 'unstructured' information, such as text, images and video. Applying these tools to political research allows political scientists to test theories in new ways or contexts. For example, through measuring legislators’ ideological positions. In the past, American political scientists developed ways of measuring legislators’ political views based on their voting records. However, this measurement didn't work in the UK. Due to the nature of the British constitution, MPs tend to vote with their party even when they strongly disagree with the party line. 

To resolve this problem, researchers developed a computational text analysis technique which statistically analyses the words that MPs use in their speeches, to detect their views on an issue. If you want to see this being used in action, researcher Tom O'Grady applies this measure to test how Labour MPs’ positions on welfare change over time. The computational measurement of text has exploded in the social sciences. Researchers who are interested in how the media reports on political events have been able to apply data science methods to huge datasets of newspaper articles. Text, however, is only one element of what media sources present. The public can also be influenced by the images which media sources use to accompany their articles. Michelle Torres has applied computer vision techniques to measure differences in the mood of the images chosen by different news sources reporting on Black Lives Matter protests in the United States.  

As we develop measures from new/unconventional data sources, we can answer more questions. Bryce Dietrich and his colleagues use audio recordings of US Supreme Court cases to measure the tone of judges’ questions posed to lawyers who are arguing a case. Remarkably, they find evidence that emotional arousal correlates with the judge’s final decision in a case. These are just two of many ways in which political scientists are applying newly developed data science techniques to answer long-standing research questions. Researchers are also using Google search histories, social media data and even public CCTV footage to gain insight into how politics, and ultimately humans, work.  

If you (or anyone you know) is thinking of applying to university, there would be no better place to start than the newly created BSc Politics and Data Science.

And lastly, do not be afraid to get stuck right in – there is an abundance of free resources for learning the basics of R or Python, and it doesn’t take much work to start doing some basic data science projects of your own!