Double Jeopardy and Climate Impact in the Use of Large Language Models: Socio-economic Disparities and Reduced Utility for Non-English Speakers

World, 2024

Reference ID

JD_SCR_001

Producer(s)

Aivin V. Solatorio, Gabriel Stefanini Vicente, Holly Krambeck, Olivier Dupriez

Metadata

JSON

Created on

Mar 24, 2025

Last modified

Mar 24, 2025

Page views

104

Overview

Abstract

This work investigates the socio-economic disparities and reduced utility for non-English speakers in the use of large language models (LLMs). We use the FLORES-200, Ethnologue, and World Development Indicators datasets to analyze the socio-economic disparities in the use of LLMs. We also use the OpenAI's GPT-4 API to assess the reduced utility of LLMs for non-English speakers.

Reproducibility Package

Scripts

Tokenization of FLORES dataset

File name

compute-premium-costs.ipynb

Title

Tokenization of FLORES dataset

Format

Jupyter Python Notebook

Description

Computes the tokenization premium for the FLORES dataset. The calculation of the population-weighted GDP for each language is also done in this notebook.

License

Name
Mozilla Public License

Back-translation task for the FLORES dataset

File name

back-translation-task.ipynb

Title

Back-translation task for the FLORES dataset

Format

Jupyter Python Notebook

Description

Generates the back-translation task for the FLORES dataset. The notebook implements the batched translation strategy for the translation task and uses the OpenAI GPT-4o API.

License

Name
Mozilla Public License

Additional analysis of the results

File name

analysis.ipynb

Title

Additional analysis of the results

Format

Jupyter Python Notebook

Description

Notebook for additional analysis of the results. Key visualizations are generated in this notebook, including the comparison of the tokenization premiums between two different tokenizers (GPT-4o vs. GPT-4 Turbo).

License

Name
Mozilla Public License

Source code repository

Repository name	Type	URI
double-jeopardy-in-llms	GitHub	https://github.com/worldbank/double-jeopardy-in-llms/tree/main

Software

Python

Name

Python

Version

3.10

Libraries

requests; pandas; docutils; jupyter-book; datasets; tiktoken; fire; openpyxl; tokenizers; ipykernel; transformers; torch; plotly; httpx; joblib; nbformat; openai; ipywidgets; groq; matplotlib; kaleido; scipy; statsmodels

Reproducibility

Technology environment

This work has been developed using a MacBook Pro with an M1 Pro processor and 64GB of RAM. No GPU is needed for the computations.

Technology requirements

Access to the OpenAI API is required.

Reproduction instructions

Some of the notebooks are not publicly available because they are used to handle proprietary data from Ethnologue which is not publicly available. One of the notebooks is used to compute the adjusted population based on the historical figures from Ethnologue and the annual population growth rates.

This repository uses poetry to manage dependencies. To install the dependencies, run the following command:
`poetry install'

To review the list of dependencies, please refer to the pyproject.toml file.

VS Code / Cursor users can use the Python extension to run the notebooks.

Use the following command to spin up a local Jupyter server:

`poetry run jupyter notebook'

It is recommended to use a virtual environment to run the code.

Additionaly, the notebooks/compute-premium-costs.ipynb notebook uses the OpenAI API. To use the API, you need to set the OPENAI_API_KEY environment variable. You can create a .env file in the root of the repository and add the following:

`OPENAI_API_KEY=<your-openai-api-key>'

Data

Datasets

FLORES-200 and FLORES+

Name

FLORES-200 and FLORES+

Note

A multilingual dataset covering 100 languages, with 1,000 sentences per language. Used for evaluating translation quality and computing the tokenization premium relative to English.

Data URL

https://github.com/facebookresearch/flores

Ethnologue

Name

Ethnologue

Note

Provides linguistic data, including the number of speakers, geographic distribution, and writing systems. We use Ethnologue to estimate the number of speakers for each language.

Data URL

https://www.ethnologue.com/

World Bank, World Development Indicators (WDI)

Name

World Bank, World Development Indicators (WDI)

Note

Contains socio-economic data at the country level. Specifically, we use the GDP per capita in current US$ (NY.GDP.PCAP.CD) and the annual population growth rates (SP.POP.GROW) indicators to compute the population-weighted GDP for each language and for aligning population estimates to 2022 based on historical figures from Ethnologue.

License

CC BY 4.0

Data URL

https://datacatalog.worldbank.org/dataset/world-development-indicators

OpenAI GPT-4o and GTP-4 Turbo APIs

Name

OpenAI GPT-4o and GTP-4 Turbo APIs

Note

Used to assess the reduced utility of LLMs for non-English speakers. We applied translation with different prompting methods to generate reference translations for FLORES sentences. The LLM translated non-English sentences into English, with the original English sentences serving as a benchmark for evaluating translation quality.

Data URL

https://openai.com/api/

Citation requirements

Please cite our paper as follows when referencing this work.

@misc{solatorio2024doublejeopardyclimateimpact,
title={Double Jeopardy and Climate Impact in the Use of Large Language Models: Socio-economic Disparities and Reduced Utility for Non-English Speakers},
author={Aivin V. Solatorio and Gabriel Stefanini Vicente and Holly Krambeck and Olivier Dupriez},
year={2024},
eprint={2410.10665},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/2410.10665},
}

Description

Output

Double Jeopardy and Climate Impact in the Use of Large Language Models: Socio-economic Disparities and Reduced Utility for Non-English Speakers

Type

Working paper

Title

Double Jeopardy and Climate Impact in the Use of Large Language Models: Socio-economic Disparities and Reduced Utility for Non-English Speakers

Authors

Aivin V. Solatorio, Gabriel Stefanini Vicente, Holly Krambeck, Olivier Dupriez

Abstract

Artificial Intelligence (AI), particularly large language models (LLMs), holds the potential to bridge language and information gaps, which can benefit the economies of developing nations. However, our analysis of FLORES-200, FLORES+, Ethnologue, and World Development Indicators data reveals that these benefits largely favor English speakers. Speakers of languages in low-income and lower-middle-income countries face higher costs when using OpenAI's GPT models via APIs because of how the system processes the input -- tokenization. Around 1.5 billion people, speaking languages primarily from lower-middle-income countries, could incur costs that are 4 to 6 times higher than those faced by English speakers. Disparities in LLM performance are significant, and tokenization in models priced per token amplifies inequalities in access, cost, and utility. Moreover, using the quality of translation tasks as a proxy measure, we show that LLMs perform poorly in low-resource languages, presenting a ``double jeopardy" of higher costs and poor performance for these users. We also discuss the direct impact of fragmentation in tokenizing low-resource languages on climate. This underscores the need for fairer algorithm development to benefit all linguistic groups.

URL

https://arxiv.org/abs/2410.10665

DOI

https://doi.org/10.48550/arXiv.2410.10665

Project website

https://github.com/worldbank/double-jeopardy-in-llms

Authoring entity

Agency Name	Affiliation
Aivin V. Solatorio	World Bank
Gabriel Stefanini Vicente	World Bank
Holly Krambeck	World Bank
Olivier Dupriez	World Bank

Date of production

2024-10

Scope and coverage

Geographic locations

Location	Code
World	WLD

Access and rights

License

Name	URI
Mozilla Public License	https://www.mozilla.org/en-US/MPL/

Information on metadata

Producers

Name
John Doe

Date of Production

2025-03-14

Back to Catalog