AI, LLMs, and Economic Decision Making
Making language-model output usable as economic analysis: every number traceable to a model, every claim traceable to a source.
- LLM agents
- RAG
- knowledge graphs
- AI-ready data
Seven lines of work, each with its own page: what the question is, what has been settled, and what is still open. Papers are listed under the project they belong to.
Making language-model output usable as economic analysis: every number traceable to a model, every claim traceable to a source.
Designing experiments that use what they have already learned — to estimate an effect precisely, or to choose the right policy, with fewer units.
Finding the best option under a fixed budget of trials — and pinning down the exact limits of how well that can be done.
Machine learning estimates of causal and structural parameters are biased by their own regularization. This project is about the correction term — how to estimate it, and what estimating it well buys you.
Estimating what an intervention does — under missing treatment labels, distribution shift between the study and the target population, or a single treated unit.
One object — the ratio of two densities — underlies covariate shift correction, learning from positive and unlabeled data, and detecting anomalies without anomaly examples.
Bringing prediction uncertainty, rather than point forecasts, into investment decisions — and applying statistical machinery to problems in financial practice.