HLER: Human-in-the-Loop Economic Research via Multi-Agent Pipelines for Empirical Discovery
TLDR
HLER is a multi-agent system for human-in-the-loop empirical economic research with dataset-aware hypothesis generation and two-loop architecture.
Reasoning
The paper introduces a novel human-in-the-loop approach that addresses the unique constraints of empirical economics, such as dataset grounding and human judgment. Its strengths include dataset-aware hypothesis generation to reduce hallucinations and a two-loop architecture with human gates. Weaknesses are its domain specificity to economics/social sciences and the abstract cut-off, which limits visibility of experimental scale and results.
Read-first score
Read-first score 64.6, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 70.
Field roles
Rank sensitivity
Stability: volatile; rank range: 27.
Keyword Scores
Deep Analysis
Innovations
- Dataset-aware hypothesis generation constrained by dataset structure, variable availability, and distributional diagnostics
- Two-loop architecture: question quality loop for screening feasible hypotheses and research revision loop for automated review-triggered re-analysis and revision
- Human decision gates embedded at key stages to guide the automated pipeline
Methodology
HLER is a multi-agent system with specialized agents for data auditing, profiling, hypothesis generation, econometric analysis, manuscript drafting, and automated review. It employs dataset-aware hypothesis generation to reduce infeasible hypotheses and a two-loop architecture with human decision gates. Experiments were conducted on three empirical datasets.
Key Results
Dataset-aware hypothesis generation produced feasible research questions in 87% of cases versus 41% unconstrained, and complete empirical manuscripts were generated at an average API cost of $0.8-$1.5 per run.