updates

Paper accepted at NeurIPS 2026: Within-Model vs Between-Prompt Variability in Large Language Models for Creative Tasks

New preprint on arXiv: Prompting Against Persona Drift: Comparing Intervention Timing and Content in LLM-Simulated Conversations

New paper in Innovation: The European Journal of Social Science Research: Teacher-centered LLM-based multi-agent systems – towards differentiated educational worksheets

New preprint on arXiv: When Does LLM Orchestration Pay Off? A Controlled Evaluation of Accuracy, Cost, and Task Difficulty

Find out more about our FACET project in the TU Berlin press release (in German)

Paper accepted at IJCAI AI and Social Good 2026: FACET: Multi-Agent AI Supporting Teachers in Scaling Differentiated Learning for Diverse Students

Meet us at the Long Night of the Sciences at the Zuse Institute Berlin: Für jeden Kopf das passende Material? Mithilfe von KI im Schulunterricht differenzieren (in German)

New preprints on arXiv:

• LLM-Based Educational Simulation: Evaluating Temporal Student Persona Stability Across ADHD Profiles
• Simulating Eating Disorder Patients with LLMs: Evaluating Psychological Persona Stability in Multi-Turn Conversations

Talk at Universitat Pompeu Fabra, Barcelona, Spain: Multi-Agent LLM Systems for Education: Simulating Diverse Learners to Support Teacher Workflows

Two posters at ACM CHI 2026, Barcelona:

• Maintaining Stable Personas? Examining Temporal Stability in LLM-Based Human Simulation
• It’s Not Just the Prompt: Model Choice Dominates LLM Creative Output

Workshop with Schulpsychologie Berlin on KI im therapeutischen Kontext (in German)

Talk at the Tag der Mathematik, Berlin: Schnell gefragt, schnell vergessen? Was KI mit unserem Lernen macht (in German)

Join our Workshop at Girls’ Day 2026: Vom Prompt zum Lernbuddy – Chatbots und KI entdecken (in German)

Panel discussion at the Weizenbaum Debate “KI gehört ins Klassenzimmer”, Weizenbaum Institute, Berlin (in German)

New role as research lead of IOL Humans and AI at the Zuse Institute Berlin, in the Interactive Optimization and Learning Lab of Sebastian Pokutta

I successfully defended my PhD thesis A Cognitive Load-Based Design Framework for Augmented Reality Training. Many thanks to Thomas Kosch and Norbert Gronau for their supervision.

New project: FACET – Framework for Agent-Based Classroom Enhancement for Teachers

New project: AI Meets Psychology and Systems