Patrick Schwabl

In my PhD project, I want to find out how artificial intelligence (AI) classifies political texts with respect to various dimensions and whether this coincides with human classifications. For this, my project is at the interdisciplinary interface between the social sciences and computer science.

I am working on different ways to compare human understanding of politics with the understanding that AI has of politics.

For example, comparisons of AI with textual data coded by humans, comparisons with survey data or expert assessments, and an evaluation of the internal coherence of AI models are suitable for this purpose.

In addition, so-called downstream tasks are common for such purposes. These evaluate the quality of the model, depending on how good it is, for example, at correctly assigning texts to the originating parties.