PhD Position F/M Multilingual and cross-cultural automatic analysis of argumentation structures in political debates

Logical fallacies in political debates are extremely present, both as errors in reasoning and as means for propaganda and spreading disinformation. Therefore, their detection has been addressed in the field of argument mining (AM). In particular, this subject falls between the realms of Natural Language Processing (NLP) - specifically AM; theoretical informatics, specifically argumentation theory; and political sciences. While the application and development of NLP techniques are necessary for the detection and the generation of arguments and their fallacies, the theory of argumentation in AI is its necessary foundation. The underlying Argumentation Framework (AF) is necessary to develop robust and explainable algorithms for the detection of argumentative fallacies. When focusing on the presence and leverage of fallacies in the political discourse, the connection to political and social sciences is evident.

In particular, this Ph.D. thesis aims to detect formal and informal logical fallacies in a multilingual corpus of political debates by integrating both argumentation theory and NLP into the detection process. Formal fallacies involve faulty logical reasoning, whereas informal fallacies may have a valid structure or form, but the premises are either untrue or irrelevant. This part of the thesis will expand on some already existing works that already focused on providing a taxonomy of formal and informal fallacies on various corpora, including political debates. The goal will also be to show that the merging of argumentation theory and NLP is a precious strategy to be leveraged for augmenting the performance of machine learning (ML) and large language models (LLMs) in detecting fallacies.