Publics, Public Opinion and Semantic Analysis

• In Event: Really Useful Analytics and the Good Life

Abstract

The emergence of semantic polling techniques (where semantic research methods are used to "read" and attribute sentiment values to large datasets, especially those generated by social media, and the results of this exercise are then used to make statements about public opinion) can be understood as a continuation of a number of pre-existing trends in public opinion research. This is because for more than two decades the orthodox paradigm of public opinion measurement has been fraying at the edges.

The rise of focus group-driven politics in the the US and the UK in 1990s injected a powerful qualitative methodology into public opinion research. More recently, declining opinion poll response rates have forced pollsters to adopt increasingly imaginative statistical techniques in an effort to make their samples more accurately representative of the electorate. Arguably this is now occurring to the point where the work of pollsters is less about polling and more about modeling (indeed, the success of Nate Silver as the ultimate election prediction expert in the 2012 US elections and his use of Bayesian-type statistical methods can be seen as the logical next step in this process).

What does this mean for how we theorise public opinion? Drawing on Susan Herbst's idea of public opinion infrastructures, this paper seeks to assess how semantic analysis fits into these broader trends, and the ethical and regulatory implications of these developments for government and politics. In particular, this paper seeks to develop a conception of the public / publics and public opinion (two closely related concepts that are not always the easiest bed fellows) that is better suited to understanding semantic analysis and the insights that it has the potential to provide. Furthermore, as new research methods change our understanding of the public and public opinion, normatively we will ask what does this mean for the idea of "good" government and democracy, and what challenges does it pose to political legitimacy when statistical models and algorithms are used to define the public’s will?

Author

• Nick Anstead, London School of Economics and Political Science