The paper offers conceptually and methodologically well-grounded approaches towards discourse analysis of people's everyday political discussions on the Internet with the aim to determine how deliberative such discussions are. The discourse ethics theory of Jurgen Habermas serves as the conceptual foundation of the study presented in the paper within his model of deliberative democracy, i.e. a democracy that advocates a need to discuss publicly different worldviews from the normative and ethical perspectives. The authors test the applicability of such an approach to online discussions focused on the politically charged topics of destroying the embargoed western food products and increasing retirement age in Russia. Over 5,000 comments posted on the discussion forums by residents of the cities of different type and size were coded and analyzed. The coding included the key deliberative features of internet-discussions. The research generates empirical evidence pointing out that the analysis of internet-discussions as online deliberative practices helps reveal certain essential aspects of people's interpretation of the publicly salient events that would be problematic to obtain through more traditional sociological methods to study social moods or computer-based text mining, such as sentiment-analysis, which do not necessarily include the moral and ethical justification of the analyzed utterances. These empirical datasets generated following the claim-based discourse-analysis were further fed, as an experiment, into the recurrent neural network in order to train it to predict positions of discourse participants in connection with the claims they made with the support of respective argumentation. The experiment demonstrates opportunities, conditions and limitations of using the artificial intelligence technologies for better understanding of public debates.