As part of our MigHANA research cooperation and in collaboration with our project partner OOWV, this master's thesis aims to explore how a Large Language Model (LLM) or generative AI such as ChatGPT can be used to formulate meaningful hypotheses regarding statistical associations or causal relationships between Key Performance Indicators (KPIs). The study will leverage the model's "common sense" and additional information about the KPIs (e.g., metadata) and domain knowledge. This thesis aims to determine what properties (content, structure, etc.) metadata and domain knowledge should have to maximize the meaningfulness of the formulated hypotheses.

• Analysis of LLM Capabilities: Investigate how LLMs like ChatGPT can be used to formulate hypotheses about statistical associations and causal relationships between KPIs.
• Integration of Metadata and Domain Knowledge: Develop methods to incorporate additional information about KPIs, such as metadata and domain knowledge, into the hypothesis formulation process.
• Hypothesis Evaluation: Assess the formulated hypotheses in terms of their meaningfulness and practical applicability.
• Optimization of Metadata and Domain Knowledge: Identify the optimal properties (content, structure, etc.) of metadata and domain knowledge to maximize the quality of the formulated hypotheses.

• Interest in artificial intelligence, particularly in generative models and machine learning
• Knowledge of data analysis and statistics
• Basic programming skills (e.g., in Python)
• Ability to understand complex relationships and think analytically
• Independent and structured working style