IV Co-Scientist: Multi-Agent LLM Framework for Causal Instrumental Variable Discovery

1. 1Identify the problem of confounding between an endogenous variable and the outcome
2. 2Use large language models (LLMs) to aid in identifying valid instruments
3. 3Implement a two-stage evaluation framework to assess the effectiveness of the IV Co-Scientist framework
4. 4Apply the framework to real-world datasets to discover causal instrumental variables

Data scientists and researchers on a team can benefit from this framework as it aids in identifying valid instruments for causal effect analysis, and ml-researchers can apply this to improve their understanding of causal relationships

💡 Large language models can aid in identifying valid instruments for causal effect analysis

🤖 IV Co-Scientist: LLM framework for causal instrumental variable discovery