AgentEconomist: An End-to-End Agentic System for Translating Economic Intuitions into Executable Computational Experiments

ACL ARR 2026 January Submission10633 Authors

Keywords: Large Language Model Agents, AI Scientist, Interactive System, Human-AI Collaboration

Abstract: A long-standing challenge in economics lies not in the lack of intuition, but in the difficulty of translating intuitive insights into verifiable research. To address this challenge, we introduce AgentEconomist, an end-to-end interactive system designed to translate abstract intuitions into executable computational experiments. Grounded in a domain-specific knowledge base covering over 8,700 high-quality academic papers, the system employs a multi-agent architecture. Specifically, an Idea Development Agent and an Experimental Design Agent collaborate to formulate theoretically grounded hypotheses and draft experimental protocols. An Experimental Execution Agent then conducts the designed experiments, forming a closed-loop workflow that supports the rapid development, validation, and iterative refinement of economic intuitions. Through extensive experiments involving human expert evaluation and large language models (LLMs) as judges, we show that the system generates research ideas with stronger literature grounding and higher novelty and insight than state-of-the-art generic LLMs. Overall, AgentEconomist adopts a human-AI collaboration paradigm that enables researchers to focus on high-level intuitions, while delegating the labor-intensive processes of translation and computational execution to agents.

Paper Type: Long

Research Area: Human-AI Interaction/Cooperation and Human-Centric NLP

Research Area Keywords: human-AI interaction/cooperation, human-in-the-loop, human-centered evaluation, user-centered design

Contribution Types: NLP engineering experiment, Publicly available software and/or pre-trained models, Data analysis

Languages Studied: English

Submission Number: 10633