An auditable AI agent loop for empirical economics

Research applicability

A working-paper workflow combining logged agent searches with post-search holdout evaluation for adaptive empirical analysis.

Getting started

Start with the working paper linked by the official summary. Predefine data splits and audit what the agent can access; record every search and reserve a held-out evaluation that cannot guide adaptation.

Maintenance

Maintenance status and current operational availability were not independently audited.

License

Not verified in this review. Consult upstream software, service, dataset and model terms before reuse.

Related resources

• EconML: causal machine-learning methods for research

Workflow & human oversight

Resource-level entry: no universal autonomy level is assigned. The researcher defines the task, controls access and validates outputs; classify an actual instantiated workflow separately.

Workflow level is not an evidence score; higher is not better evidence.

Approach

The official summary describes adapting an open-source agent-loop architecture to forecast combination, logging the search and evaluating on held-out data.

Limitations

Only the official working-paper summary was reviewed, not the full paper, code or experiments. The summary reports benchmark improvements by the authors; no magnitude, independent reproduction or general research-efficiency benefit is established here.

Reproducibility

Reviewed stored primary-source extract candidate-evidence-048.md (retrieved 2026-10-15). Documentation review only; no downloaded code, skill or model was executed.

Adoption requirements

A research question, reproducible data/code, logged searches and a genuine untouched evaluation set. Full paper and implementation need review before adoption.