Overview

Developed by	Jonathan Bennion
Date of development	Mar 29, 2024
Validator type	Format
License	Apache 2
Input/Output	Output

Description

This bias check format validator ensures textual outputs do not contain biased language towards specific demographics, such as race, gender, sex, religion, ethnicity.

Intended Use

This validator can be used to ensure fairness of model output across various demographic groups.

Requirements

• Dependencies:

  • guardrails-ai>=0.5.0
  • transformers>=4.40

• Dev Dependencies:

  • pytest
  • pyright
  • ruff

• Foundation model access keys:

  • None (basic HuggingFace hub access)

Installation

$ guardrails hub install hub://guardrails/bias_check

Usage Examples

Validating string output via Python

In this example, we apply the validator to a string output generated by an LLM.

# Import Guard and Validator
from guardrails.hub import BiasCheck
from guardrails import Guard

# Setup Guard
guard = Guard().use(
    BiasCheck(threshold=0.9, on_fail="exception")
)

guard.validate("The movie was great!") # Validator passes
guard.validate("Why do men always think the movie was great?")  # Validator fails