About

The Project

The Global Representativeness Index (GRI) was developed to address a gap in survey methodology: while response rates and sample sizes are routinely reported, there has been no standardized, quantitative measure of how well a sample represents the global population across multiple demographic dimensions.

As large-scale surveys and public consultations increasingly shape AI policy and governance, the need for transparent representativeness measurement has become critical. The GRI provides researchers, policymakers, and survey practitioners with a rigorous, open-source tool for this purpose.

Authors

Evan Hadfield The Collective Intelligence Project

Andrew Konya Remesh.ai

Citation

If you use the GRI in your research, please cite:

@software{hadfield2025gri,
  author    = {Hadfield, Evan and Konya, Andrew},
  title     = {The Global Representativeness Index: Measuring Demographic
               Representativeness in Survey Samples Using Total Variation Distance},
  year      = {2025},
  url       = {https://github.com/collect-intel/gri},
  license   = {MIT}
}

An arXiv preprint is forthcoming.

License

This project is released under the MIT License. You are free to use, modify, and distribute the code and methodology with attribution.

Data Sources

The GRI benchmarks are derived from the following authoritative datasets:

Dataset Source Year Description
World Population Prospects UN DESA Population Division 2023 Population by country, gender, and 5-year age group
Global Religious Landscape Pew Research Center 2010 Religious composition by country
World Urbanization Prospects UN DESA Population Division 2018 Urban/rural population distribution by country

All benchmark data files and full source attribution are available in the repository under data/raw/benchmark_data/.

Contributing

We welcome contributions. Please open an issue on GitHub for bug reports, feature requests, or questions about the methodology.

Contact

For inquiries about the GRI project or the Global Dialogues survey series, please reach out through GitHub Issues.