Towards the Automated Measurement of Deliberative Communication

Deliberation refers at the macro-level to deliberative democracy, which in taking collective decisions relies on consensus-oriented and argumentative communication instead of majority decision-making. At the micro-level, it refers to specific deliberative fora, set up to deal with a given decision problem, be it the adoption of a collective rule or the resolution of a local conflict. The idea of deliberation requires at both levels that certain attributes of the institutional setting, the communication and the individual behavior be present. In this paper we derive a definition of deliberation based on ten dimensions including institutional, communicative and individual properties. Further we focus on those dimensions on which the actual communication in a discourse can contribute information. In this respect, our approach is very similar to the DQI (e.g. Steenbergen et al. 2009). In a next step, we develop, in cooperation with linguists and information scientists, measurement concepts that can be automatically applied to empirical dialogues. The paper aims at showing how far we can already get using automated text analysis and to which extent this can contribute to measuring the deliberative quality of dialogues. Our measures include shallow statistical indicators based on stemming and frequency, as well as deep linguistic analysis. For instance, we use conceptual recurrence plots (Angus et al. 2011) to detect patterns of interactions across the deliberative discourse. In combination with sentiment analysis (see e.g. Wanner et al. 2009), we automatically identify and extract subjective stances towards other participants and their (counter)arguments to approach an automated measure of the compelling power of arguments. The empirical materials used include the digitalized version of the mediation on “Stuttgart 21”, a hot conflict about the reconstruction of the central station in Stuttgart, Germany, and dialogues taken from role-play negotiation experiments.