From RC Trust intern to PhD researcher

How do people adjust their behavior when they interact with artificial intelligence – and how does AI, in turn, anticipate human reasoning? This question was at the heart of Nadya Hanaveriesa’s internship at the Research Center Trustworthy Data Science and Security.
From July to September, she joined the Human Understanding of Algorithms and Machines (HU[AM]) group, supervised by Prof. Nils Köbis and Dr. Inês Terrucha. The group studies human and machine behavior and how people understand and interact with AI systems.
Nadya recently completed her master’s degree in psychology at New York University, specializing in experimental and social psychology. Her master’s thesis focused on human-AI cooperation and social norms – a connection that drew her to HU[AM].

Looking at both sides

During her internship, Nadya worked on strategic reasoning in human-AI interaction. She examined how perceived reasoning ability can influence behavior depending on whether the interaction partner is another person or an AI system. Her approach reflected her interest in both sides of that interaction. Rather than running another experiment with human participants, she worked with large language models as “participants,” prompting different models via APIs and comparing how they behaved toward humans. The results revealed an intriguing pattern. People engaged in deeper strategic reasoning when interacting with AI than with other humans, while LLM agents showed a similar pattern. A mismatch emerged, however, in how several models appeared to estimate how deeply humans would reason – estimates that did not always match actual human behavior.

From internship to PhD

For Nadya, the project also meant stepping beyond familiar methods. Coming from psychology, she learned programming and worked directly with LLMs. One insight stayed with her: as AI increasingly becomes an active part of an interaction rather than simply a tool, studying human behavior alone is no longer enough. Understanding human-AI interaction also means examining how AI behaves and how it represents or anticipates human behavior. Especially when each side’s decisions depend on what it expects the other to do.
She particularly valued HU[AM]’s open research culture and its focus on learning and progress. The internship also helped clarify her next steps: Nadya says she has become much more certain about the research questions she wants to pursue. That next step is now taking shape: in December 2026, she will return to Prof. Nils Köbis’ HU[AM] group as a PhD researcher.

Internship Program

The RC Trust Internship Program gives researchers from around the world the opportunity to join interdisciplinary research on trustworthy AI. Interns work closely with RC Trust research groups, while living expenses, housing, and travel costs are funded.
Learn more about the RC Trust Internship Program.