A Modal View on Polder Politics

The media plays an increasing role in the functioning of our society, often being the main informant of citizens and policy-makers. Controversies easily arise when the media are accused of inaccurate or biased information provision, such as the affair around the BBC and the suicide of David Kelly in the UK and the alleged demonisation of Pim Fortuyn before his assassination in the Netherlands. The messages of politicians reach most citizens through the media rather than directly, making the image of a party or candidate and their reported positions on key issues an important factor in election campaigns. Therefore, it is important to analyse the content of the media and its effects on media consumers, such as it is often done in election studies, agenda setting, and framing research [Bryant and Zillman, 2002, Goffman, 1974, McCombs and Shaw, 1972]. These studies generally annotate news items using Content Analysis which has resulted in a large number of annotated articles over the past decades. A uniform way to store and query this media data can lead to easier and more accurate (meta-)research. Moreover, this can allow general purpose analysis and visualisation tools to be developed. In this paper we provide such a uniform representation by formalising media data as Kripke structures. Not only do we get unified representations with formal semantics, we also inherit well studied query languages in the form of modal logics. This allows us to use off-the-shelf model checking software to query newspaper articles, or at least the formalisation of these articles. In order to reproduce recent results of a media study about the Dutch elections in 2002 [Kleinnijenhuis et al., 2003], we show that quite expressive modal languages (such as Hybrid Logic H(@, ↓)) are needed to capture important political terms, such as internal debate or criticism. In our experiments, we take almost 7500 newspaper articles, and transform them into Kripke structures with around 35.000 edges. Using an