Seminar Data-Science II (Empirical Studies)

The learning goals will be achieved by in-depth learning from examples on the basis of personally assigned scientific topics as well as by active participation in the presentation sessions. The organizer will choose subject areas within the field of data science and provide scientific papers for the students to work through.

Previous participation in courses offered by our chair is recommended.

On the basis of scientific literature, in particular original research articles, students independently familiarize themselves with one topic in data science. They classify and narrow down the scope of the topic appropriately and provide a critical evaluation. Students work out concepts, procedures and results of a given topic clearly and with appropriate formalisms. They do so in-depth by writing in a concise manner and to a defined extent. They are expected to show evidence of independent development of their arguments by presenting self-selected examples. Towards the end of the course they give a descriptive, in-depth oral presentation of the data science topic using suitable media and examples in an agreed-on format.

This seminar will be split into two main topic blocks. Every student will be assigned a research paper from only one of these blocks to work on. Yet, it is expected that students also actively participate in discussion on papers from the other topic blocks after they have been presented.

When applying for this seminar, please indicate whether you would be interested in only one or both topic blocks. The two topics we are going to discuss in the FSS 2025 are:

1. Large Language Models in Politics. Large Language Models can analyze ideologies, predict politicians’  views, mimic public figures, and even simulate debates. But they’re not perfect. For example, biases and ethical concerns are big challenges. We will explore how these models currently are utilized to analyze the political landscape.
   • Llama meets EU: Investigating the European political spectrum through the lens of LLMs
   • Large Language Models can impersonate politicians and other public
   • Large Language Models Can Be Used to Estimate the Latent Positions of Politicians
   • Systematic Biases in LLM Simulations of Debates
   • Out of One, Many: Using Language Models to Simulate Human Samples
   • Hidden Persuaders: LLMs’ Political Leaning and Their Influence on Voters

2. Simulating Humans with LLMs. Role-playing or simulation of synthetic personas in LLMs involves designing and deploying models that emulate specific characters, professions, or cultural perspectives to achieve targeted interactions. However, it also raises significant ethical and technical questions, such as ensuring authenticity and avoiding harmful stereotypes. We will explore how synthetic personas can enrich user interactions and social science research while balancing realism, and ethical considerations in deploying LLMs in diverse scenario.

   • Are Personalized Stochastic Parrots More Dangerous? Evaluating Persona Biases in Dialogue Systems
   • Assessing Generalization for Subpopulation Representative Modeling via In-Context Learning
   • Bias Runs Deep: Implicit Reasoning Biases in Persona-Assigned LLMs
   • Evaluating Cultural Adaptability of a Large Language Model via Simulation of Synthetic Personas
   • Generative language models exhibit social identity biases
   • On the steerability of large language models toward data-driven personas​​​​​​​

If you are interested in this seminar, please apply to Maximilian Kreutner via email. 

Please start the Subject Line with “[SemDSII]”, and provide some details about your background, e.g., whether you have taken some relevant classes before (cf. course description).