Argument and debate form cornerstones of civilised society and of intellectual life. Argument Mining (AM) is a multidisciplinary research field, encompassing diverse areas such as computer sciences, logic and philosophy, language and rhetoric, and only recently opening up to the social sciences. In algorithmically-assisted AM several artificial Intelligence-based techniques are used, i.e.: natural language processing (NLP); semantic and logical analysis; deep learning; Transformer-based models; clustering topic-dependent arguments. These techniques allow to extract arguments from generic textual corpora, in order to provide structured data for computational models of argument and reasoning engines.
Unfortunately, from the social sciences point of view, these new AM techniques, despite their great potential, are often applied too narrowly. That means they are limited mainly to the scope of interest of computer science and linguistic issues, without catching wider social context, and examining mainly the persuasiveness of a particular argument, without examining its location on the debate map. Public sphere theories underscore the critical role of entire debates, not merely individual arguments, in influencing civic mindsets, engaging citizens in community activities, and fostering social bonds. To capture this, we propose a new model exploring the possibility of using the AI-based AM to extract argumentative structures from debates in their whole social complexity. In the previous research the AM methods have not been yet practically connected with the theoretical models of civic participation, taking into account both the achievements of social science and computer science.
The debate space that will be explored in this project will be the online public sphere. Although some researchers consider deliberative model (credited to J. Habermas, J. Rawls, J. Cohen, J. Fishkin etc), as the most appropriate normative theory for modelling rational debate on the Internet, we hypothesize that it is not the only possible and necessary one. The agonistic model (based on works of C. Mouffe, E. Laclau etc) will be considered as an alternative way to describe public debates. Our goal will be to parameterize and measure the features of these models in online communication, taking into consideration the structure of arguments and their role in the debate. The opposite of these models is the antagonistic “anti-model”, characteristic of toxic communication, when arguments are lost under the pressure of negative emotions and fake news & misinformation. This anti-model will also be characterised in the context of improper argumentation used there, and presented as a negative benchmark.
The project is primarily focused on research objectives in the field of social sciences, but due to its interdisciplinary nature, it also includes research objectives relevant to computer sciences. In particular, in this project we will introduce a novel approach to AM (“Hybrid AM Model”), which incorporates a hybrid combination of two methods:
A long-term objective is building a new, multilingual text corpus concerning most current public debate topics, including English and Polish databases. This corpus, collected and annotated to use for AI-trained AM, will contain data on the most current topics of public debate that will emerge during project’s work. The accumulated knowledge on argumentation patterns will be presented in knowledge graphs and formal ontologies.
The study will provide answers for the following Research Questions: