Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

LLMs Model Non-WEIRD Populations: Experiments with Synthetic Cultural Agents

Augusto Gonzalez-Bonorino, Monica Capra, Emilio Pantoja

TL;DR

This work introduces Synthetic Cultural Agents (SCAs) that operationalize non-WEIRD populations within economic experiments, addressing WEIRD bias in behavioral research. By grounding SCAs with Knowledge Bases built via a Search + Retrieval-Augmented Generation pipeline and prompting LLMs to act as tribe members, the authors pilot dictator, ultimatum, and multimodal endowment-effect experiments across six small-scale societies. The results reveal substantial cross-cultural variation, with none of the SCAs behaving as purely self-interested agents and notable patterns aligning with anthropological expectations, including Yanomami profiles that approach Homo economicus in some roles. The study argues that SCAs provide a flexible, ethical, and cost-effective means to generate hypotheses and pilot protocols for hard-to-reach populations, while acknowledging limitations and outlining avenues for refinement and prospective field validation.

Abstract

Despite its importance, studying economic behavior across diverse, non-WEIRD (Western, Educated, Industrialized, Rich, and Democratic) populations presents significant challenges. We address this issue by introducing a novel methodology that uses Large Language Models (LLMs) to create synthetic cultural agents (SCAs) representing these populations. We subject these SCAs to classic behavioral experiments, including the dictator and ultimatum games. Our results demonstrate substantial cross-cultural variability in experimental behavior. Notably, for populations with available data, SCAs' behaviors qualitatively resemble those of real human subjects. For unstudied populations, our method can generate novel, testable hypotheses about economic behavior. By integrating AI into experimental economics, this approach offers an effective and ethical method to pilot experiments and refine protocols for hard-to-reach populations. Our study provides a new tool for cross-cultural economic studies and demonstrates how LLMs can help experimental behavioral research.

LLMs Model Non-WEIRD Populations: Experiments with Synthetic Cultural Agents

TL;DR

This work introduces Synthetic Cultural Agents (SCAs) that operationalize non-WEIRD populations within economic experiments, addressing WEIRD bias in behavioral research. By grounding SCAs with Knowledge Bases built via a Search + Retrieval-Augmented Generation pipeline and prompting LLMs to act as tribe members, the authors pilot dictator, ultimatum, and multimodal endowment-effect experiments across six small-scale societies. The results reveal substantial cross-cultural variation, with none of the SCAs behaving as purely self-interested agents and notable patterns aligning with anthropological expectations, including Yanomami profiles that approach Homo economicus in some roles. The study argues that SCAs provide a flexible, ethical, and cost-effective means to generate hypotheses and pilot protocols for hard-to-reach populations, while acknowledging limitations and outlining avenues for refinement and prospective field validation.

Abstract

Despite its importance, studying economic behavior across diverse, non-WEIRD (Western, Educated, Industrialized, Rich, and Democratic) populations presents significant challenges. We address this issue by introducing a novel methodology that uses Large Language Models (LLMs) to create synthetic cultural agents (SCAs) representing these populations. We subject these SCAs to classic behavioral experiments, including the dictator and ultimatum games. Our results demonstrate substantial cross-cultural variability in experimental behavior. Notably, for populations with available data, SCAs' behaviors qualitatively resemble those of real human subjects. For unstudied populations, our method can generate novel, testable hypotheses about economic behavior. By integrating AI into experimental economics, this approach offers an effective and ethical method to pilot experiments and refine protocols for hard-to-reach populations. Our study provides a new tool for cross-cultural economic studies and demonstrates how LLMs can help experimental behavioral research.
Paper Structure (35 sections, 6 figures, 13 tables)

This paper contains 35 sections, 6 figures, 13 tables.

Table of Contents

  1. Introduction
  2. Methods
  3. Building the Knowledge Base and Cultural Profile
  4. Search + RAG Methodology
  5. Creating Synthetic Cultural Agents and Running Experiments
  6. Selection of Small-scale Societies
  7. Instantiating Synthetic Cultural Agents
  8. Experimental Setup
  9. Results
  10. Dictator Game
  11. Ultimatum Game
  12. Key Findings from Dictator and Ultimatum Games
  13. Endowment Effect: A Multimodal Approach
  14. Discussion
  15. Methods
  16. ...and 20 more sections

Figures (6)

  • Figure 1: Framework for Constructing Experiments with Synthetic Cultural Agents
  • Figure 2: Dictator's Count to Range of Proposed Splits by Tribe and ChatGPT
  • Figure 3: Ultimatum Game Counts to Range of Offers by Tribe and ChatGPT
  • Figure A1: RAG + Search methodology
  • Figure A2: Interactive Platform: Choice of Items for the Endowment Effect with the Aché
  • ...and 1 more figures