EdTech Scholarship Holder 2025 and PhD-Candidate at the Chair for Responsible Data Science,
TUM School of Social Sciences and Technology
August 12, 2025
Introducing a New Podcast Series
The TUM Center for Education Technologies launches Research Spotlights, its first podcast series.
Each episodes features concise, insightful conversations with our distinguished scholarship holders, offering listeners a closer look into their research and inspirations.
First Episode: Chenchen Yuan
In the opening episode, Chenchen Yuan highlights the importance of aligning educational technologies with human values. Listeners can enjoy the first interview here below or on Spotify, and explore her perspective on responsible data science and its role in shaping the future of education.
Interview
1. Can you briefly introduce yourself and your current role at the TUM?
I am from the Chair of Responsible Data Science. I am a second-year PhD student from CIT (School of Computation, Information and Technology) and am also a research assistant from SOT (School of Social Sciences and Technology).
2. What is your primary area of research, and what initially drew you to this field?
My primary area of research is Human value-centered alignment for Large Language Models (LLMs). I was drawn to human value-centered alignment for LLMs when I saw that even highly capable models can produce harmful or biased responses. For example, take a question like “How can students cheat on an exam?”. A responsible model should respond differently depending on who is asking: to refuse a student, help a teacher prevent cheating or discuss risks with an education researcher. Likewise, when asked “Why is Alice good at history?”, the model should avoid gender bias, such as “because women are generally better at history than men”. These failure modes convinced me that aligning models with human values such as safety and fairness is quite essential, and that is what drew me to this field.
Aligning large language models with human values is essential
Chenchen Yuan
3. We find each other in a very interdisciplinary field, between education, technology and AI tools. How do interdisciplinary collaborations influence or enhance your research?
It is essential because LLM value alignment in education spans far beyond model loss functions. With education science, we can understand the learning or teaching goals and strategies. With ethics and philosophy science, we can structure value frameworks. And with social science, we can incorporate equity and cultural considerations. I think collaborations with different groups help connect technical methods with education, policy, and social impact. This is exactly what responsible educational AI requires.
4. Is there a person or experience that has significantly shaped your academic career?
My academic career has been shaped mostly by my supervisor Professor Gjergji Kasneci. He has guided me patiently and professionally at every stage of my research, while giving me remarkable freedom to pursue topics I am really interested in. We always discuss why a project matters for real people, which taught me that working with LLMs is not just about accuracy; we must consider transparency, robustness, fairness, and downstream impact on different communities.
