We are currently looking for a bright Master's student interested in applying artificial intelligence to simulate worker behavior for decision-making in an industrial safety context, for a cooperation between the chair for the Theory of Neural Systems and the Department of Work and Organizational Psychology.
Ultimately, this should lead to a meaningful model for rule design and reward framing in real workplaces.
The main focus of the thesis will be the evaluation of reinforcement learning techniques for a provided industry simulation. That includes implementation in the Python programming language.
The student becomes acquainted with the theoretical framework of reinforcement learning (RL), then learns to phrase the given simulation in it, and understands the conceptual limitations of such an approach.
The student implements and experiments with different RL-algorithms and explores means to robustly achieve close-to-optimal behavior in the simulation (and possibly variants). This might include overcoming technical difficulties in interfacing with the current implementation of the framework.
A thorough experimental study is conducted to investigate the role of rule design in industrial safety. The study is then interpreted from the perspective of work psychology. The "Department of Work and Organizational Psychology" will guide this evaluation, no working knowledge of the topic is required.
The specifics of the content might change, depending on intermediate results in Phase 2.
- previous exposure to machine learning techniques, e.g., by successful (80%+) participation in one or more of the INI's lectures on the topic
- good command of Python or another programming language
- no fear in the face of applied math
- psychology background is NOT required for this thesis