Panagiotis Fytas

PhD thesis: Extracting Insights into Social Sciences by Combining Natural Language Processing with Causal Inference

Machine Learning (ML), along with Explainable AI, possess the ability to discover knowledge from natural language: explaining the predictions of a text classifier provides insights into a data-generating process. Still, interpreting the explanations of classifiers causally is unwarranted since, in general, ML models detect correlations. Nevertheless, in recent years there has been a growing interest in enabling ML and even Natural Language Processing (NLP) to adjust for Causal Inference. My PhD focuses on implementing a framework that reconciles Natural Language Processing with Explainable AI and Causal Inference. The overarching objective is to utilise this framework to extract knowledge about various social science domains and, ultimately, garner an understanding of the underlying human behaviour.

Who or what inspired you to pursue your research interests?

During my undergraduate degree in Electrical and Computer Engineering, I became fascinated with computer science. Recognising Machine Learning as essential knowledge for a Computer Scientist, I decided to pursue the MSc Advanced Computing degree in Imperial College London, where, as part of my master thesis, I got to work with NLP and Explainable AI. Since then, I have discovered that carrying out research in the field of NLP is my ideal career path.