Thesis Topics

I am supervising both Bachelor and Master Theses at the Johannes Gutenberg University. You are welcome to write in either German or English.

Overall, I am interested in the topics of Human-Robot and Human-AI Interaction, Large Language Models, Dialogue Systems and Robotics. If you have a cool idea for a thesis topic in one of these areas yourself, don't hesitate to reach out! If you are still looking for an inspiration, have a look at potential topics below.

This project is in collaboration with TU/e in the Netherlands

  1. Using multimodal Vision Language Models to classify human actions.
    In this project, we would like to understand the intention that people have when they are approaching robots with different tools. We are collecting data for this project in spring 2026. Your task is to research and compare different models to classify what object people have in their hands, and together with the language they are using whether they have a positive or negative intent towards the robot.

  2. Training and developing two personalities for the robot, e.g. one that is shy and one that is sassy.
    In this thesis topic, you use a machine learning algorithm (for example Reinforcement Learning) to let the robot behave coherently and consistently in different personalities.

To Care or to Harm - Human Actions in Seemingly Unsupervised Interactions with Robots

  1. Using a dataset of people describing abstract shapes as a new benchmark for large language models and/or vision language models.

  2. Evaluation of AI Personalities, with a focus on coherence and consistency over time as well as how expectable the AI behavior is given the personality. This thesis should focus on developing ways to automatically judge whether an AI stays in a given role / personality, especially within longer conversations.

  3. Developing a dashboard to monitor the health / conversational quality of an AI agent.

Benchmarking & LLM Evaluation

  1. Match a human personality / the preference a human has for their AI chatbot with the way a LLM generates output. This thesis project should contain the development of a continuous spectrum to display personality. So instead of a binary (e.g. shy vs extroverted), the human should be able to ask the LLM to take on nuanced personality-driven behavior.

AI & Personality

LLMs & their applications

I am always interested in the development of applications in which large language models really make sense and can benefit society. If you do choose a thesis topic in this area, the thesis should not only contain the development of the application but also a user study.

  1. Can LLMs help doctors to translate accurate medical language about a diagnosis or treatment to a language that is better understandable for patients?

  1. Using thematic analysis on large datasets to extract information regarding how people feel about a topic (datasets could, for example, be Reddit or YouTube comments).

  2. Would people write different language for a robot that presents feminine (uses "she" as pronoun) vs a robot that presents masculine (uses "he" as pronoun) vs a mechanical robot (uses "it" as pronoun). This thesis work entails collecting data in a user study online and then analyzing patterns in the collected data.

Bias & Ethics