Understanding policy ideas and policy decisions: Can using big data help?

Many theories of the policy process include an important role for ideas, against a background of many other factors. In this paper, we ask whether big data can be used to better understand policy ideas and their role in the policy process. This is examined theoretically and empirically using ‘innovation’ policy as a case study. Policy debate occurs within a context of institutions, events, public opinion, and general economic and social trends (context and events). The paper privileges ‘debate’ in this examination, so ideation – the process of generating, developing, and communicating ideas - is the key focus. Our framework consists of three interrelated components: 1) the actors involved (individuals and institutions); 2) their framing of policy problems and solutions (ideas and how they are defined and presented); and 3) the forums in which discussion occurs (the institutions, networks and subsystems where actors promote their ideas). All of these are of course interrelated, but it is the framing/ideation that is the focal point in this paper. These three elements, and their interrelationships, change over time, either because of deliberate attempts to include a different range of actors, to change the framing of problems and solutions, or to shift the venue, or because of some exogenous force (or any combination thereof). At certain points in time, the changed debate might result in policy change in the form of decision-making. The questions for this paper to answer are: ▪️ What major ideas/framings can be gleaned from documentary evidence and analysed using natural language processing (NLP)? ▪️ Can these ideas be linked to policy decision-making using publicly available documents? ▪️ What does NLP add to understanding policy ideas and policy decision-making? The explosion of digital sources of information and developments in computational approaches for harvesting data on these relationships provide major new opportunities for researching this topic, providing ‘big data’ that was previously not available. The digitisation of government documents, the mass media, and the scientific literature offers enormous potential to harness these data to understand major policy ideas, how they change over time, and their relationship (if any) to government policy decisions. This paper will present a more developed outline of our theory and empirical analyses of ‘innovation’ policy as a case study to examine how useful NLP is for answering policy debate and decision questions. It will discuss some of the benefits and challenges in using ‘big data’ to answer three different analyses of innovation policy and how it changes over time: 1) the meaning of innovation in policy documents in Australia; 2) the fit of innovation definitions in Australia in relation to international trends represented by the OECD; and 3) the correspondence of meanings of innovation across expert, policy and media sources. The aim is to provide some answers on whether big data and NLP can help advance an understanding of these aspects of the policy process.