Methods to Computational Social Sciences (CSS)

Social media generates new types of data for quantifying human behavior, but also raises interesting research questions about the interaction between technology and society. In this course, we will learn the basics of computational social science, with a focus on text-as-data, and apply its methods to real-world data science problems from social media and other sources. 

The lecture will introduce you to social science research questions and findings, while the exercise will give you the practical skills to answer social questions yourself using computational tools. The first part focuses on basic techniques of text analysis. We will learn how to collect text data from sources such as newspapers, parliamentary minutes or social media, and how to pre-process it for natural language processing (NLP), such as sentiment analysis. 

We will then move on to analyzing the structural factors of online data, from time series to networks and their dynamics. We will also explore how to collect such data from social media and how to conduct online field experiments. We will then have the tools to move on to applied research questions, covering topics such as political polarization, misinformation, public health and human mobility. We will learn what conclusions can and cannot be drawn from these descriptions. Finally, we will go beyond description and learn about causality, simulation models of social systems that allow mechanistic understanding. 

The last part of the course will focus on NLP again, this time with a focus on deep learning methods such as Large Language Models (LLMs). We will use transformer models to identify complex patterns in the communication of German parties on social media. 

In summary, this course will provide a toolbox for describing different types of social systems with a focus on textual data, as well as a family of modelling approaches that can go beyond individual-level statistics to describe collective processes.