Digital Data and Computational Methods

Digital platforms and artificial intelligence have created unprecedented opportunities for studying society at scale. At the same time, digital data are not neutral representations of human behavior. What researchers can observe depends on platform architectures, algorithms, APIs, sampling mechanisms, and increasingly on automated systems that collect, classify, and interpret data.

Our research investigates these methodological challenges and develops computational approaches for collecting and analyzing large-scale digital data. A particular focus is on data quality, completeness, representativeness, and reproducibility. Our work has examined how social media APIs shape the data available to researchers and developed approaches for obtaining more systematic and complete samples from platforms such as Twitter and TikTok.

We also study the opportunities and limitations of Large Language Models as research instruments. This includes their use for text annotation, qualitative coding, classification, and the analysis of complex social and political communication. Rather than treating AI as a black-box replacement for established research methods, we investigate when these systems produce valid and reliable results, how they compare with human experts, and where new sources of bias and uncertainty emerge.

This work reflects a broader principle of Computational Social Science: the tools and data through which we observe society must themselves be objects of scientific inquiry.