Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

LangFair: A Python Package for Assessing Bias and Fairness in Large Language Model Use Cases

Dylan Bouchard, Mohit Singh Chauhan, David Skarbrevik, Viren Bajaj, Zeya Ahmad

TL;DR

Large Language Models exhibit bias in prompt-dependent contexts, making model-wide metrics insufficient for real-world safety and fairness. LangFair provides a prompt-aware, output-based evaluation framework that accommodates task-specific risks through a BYOP approach. It offers dataset generation utilities (ResponseGenerator, CounterfactualGenerator) and a modular metrics suite spanning toxicity, stereotypes, counterfactual fairness, and classification, plus AutoEval for semi-automated assessments. These capabilities support governance audits and practical fairness assessments in diverse LLM use cases.

Abstract

Large Language Models (LLMs) have been observed to exhibit bias in numerous ways, potentially creating or worsening outcomes for specific groups identified by protected attributes such as sex, race, sexual orientation, or age. To help address this gap, we introduce LangFair, an open-source Python package that aims to equip LLM practitioners with the tools to evaluate bias and fairness risks relevant to their specific use cases. The package offers functionality to easily generate evaluation datasets, comprised of LLM responses to use-case-specific prompts, and subsequently calculate applicable metrics for the practitioner's use case. To guide in metric selection, LangFair offers an actionable decision framework.

LangFair: A Python Package for Assessing Bias and Fairness in Large Language Model Use Cases

TL;DR

Large Language Models exhibit bias in prompt-dependent contexts, making model-wide metrics insufficient for real-world safety and fairness. LangFair provides a prompt-aware, output-based evaluation framework that accommodates task-specific risks through a BYOP approach. It offers dataset generation utilities (ResponseGenerator, CounterfactualGenerator) and a modular metrics suite spanning toxicity, stereotypes, counterfactual fairness, and classification, plus AutoEval for semi-automated assessments. These capabilities support governance audits and practical fairness assessments in diverse LLM use cases.

Abstract

Large Language Models (LLMs) have been observed to exhibit bias in numerous ways, potentially creating or worsening outcomes for specific groups identified by protected attributes such as sex, race, sexual orientation, or age. To help address this gap, we introduce LangFair, an open-source Python package that aims to equip LLM practitioners with the tools to evaluate bias and fairness risks relevant to their specific use cases. The package offers functionality to easily generate evaluation datasets, comprised of LLM responses to use-case-specific prompts, and subsequently calculate applicable metrics for the practitioner's use case. To guide in metric selection, LangFair offers an actionable decision framework.
Paper Structure (13 sections, 1 figure, 1 table)

This paper contains 13 sections, 1 figure, 1 table.

Table of Contents

  1. Introduction
  2. Generation of Evaluation Datasets
  3. ResponseGenerator class.
  4. CounterfactualGenerator class.
  5. Bias and Fairness Evaluations for Focused Use Cases
  6. Toxicity Metrics
  7. Stereotype Metrics
  8. Counterfactual Fairness Metrics for Text Generation
  9. Counterfactual Fairness Metrics for Recommendation
  10. Fairness Metrics for Classification
  11. Semi-Automated Evaluation
  12. AutoEval class.
  13. Conclusions

Figures (1)

  • Figure 1: Flowchart of internal design of AutoEval.evaluate method.