Research Associate for the Project “Evaluating, Explaining, and Enabling Ethical Multi-Agent Systems of Large Language Models (E4-MALM)” § 28 Subsection 3 HmbHG

Duties include academic services in the project named above. Research associates may also pursue further academic qualifications outside of their work responsibilities. They may also pursue doctoral studies outside of working duties.

The position is located at the Professorship of Trustworthy AI that focuses on the computational processing of unstructured data, mainly text. Our goal is to conduct holistic research in artificial intelligence: while most research in natural language processing and artificial intelligence focuses on obtaining state-of-the-art results on established benchmark data sets only, we seek to couple research on improving the effectiveness of our systems with the consideration of ethical aspects. That is, we aim to develop effective but also environmentally sustainable, fair, and inclusive solutions.

The employee will be hired within the project “Evaluating, Explaining, and Enabling Ethical Multi-Agent Systems of Large Language Models (E4-MALM)”, which is funded by the German Research Foundation (DFG) as part of the Priority Program LaSTing (SPP 2556) on Language Science and Technology. The aim of this project is to advance the safety and ethical reliability of multi-agent systems of large language models (MALMs). Concretely, E4-MALM pursues three core objectives: (1) developing robust evaluation frameworks and stress-testing methods to assess ethical and emergent behaviors in MALMs, (2) explaining collective failure modes (e.g., toxic agreement, groupthink) at a mechanistic level using interpretability methods, and (3) enabling ethical behavior through parameter-efficient fine-tuning and activation steering techniques guided by these mechanistic insights.

Within this project, the employee is highly encouraged to conduct a Ph.D. covering fundamental research in natural language processing and artificial intelligence with a focus on one or multiple of the following topics:

• Large Language Models
• Multi-Agent Systems of LLMs
• Mechanistic Interpretability
• Ethical Artificial Intelligence (Fairness, Sustainability, Inclusion, etc.)

The research should result in international top-tier publications (conference or journal articles) in natural language processing (e.g., ACL), artificial intelligence (e.g., AAAI), and/or interdisciplinary venues. The supervisors of the Ph.D. candidate will be Prof. Dr. Anne Lauscher and Dr. Jae Hee Lee. Both will be available for questions related to the position via email.

Our research group is dynamic, ambitious, and nationally as well as internationally very well connected. Diversity and the support of individuals identifying as members of marginalized groups are important to us.