Runtime Monitoring and Enforcement of Conditional Fairness in Generative AIs

TL;DR

This paper introduces methods for runtime monitoring and enforcement of conditional fairness in generative AI, addressing fairness concerns specific to broad, context-dependent outputs and proposing techniques to ensure fairness thresholds are maintained.

Contribution

It presents novel characterization and enforcement techniques for fairness in GenAI, including bounding worst-case unfairness and a prompt injection scheme for enforcement.

Findings

Effective enforcement of fairness thresholds in GenAI models

Validation of methods on state-of-the-art systems

Introduction of combinatorial testing for intersectional fairness

Abstract

The deployment of generative AI (GenAI) models raises significant fairness concerns, addressed in this paper through novel characterization and enforcement techniques specific to GenAI. Unlike standard AI performing specific tasks, GenAI's broad functionality requires ``conditional fairness'' tailored to the context being generated, such as demographic fairness in generating images of poor people versus successful business leaders. We define two fairness levels: the first evaluates fairness in generated outputs, independent of prompts and models; the second assesses inherent fairness with neutral prompts. Given the complexity of GenAI and challenges in fairness specifications, we focus on bounding the worst case, considering a GenAI system unfair if the distance between appearances of a specific group exceeds preset thresholds. We also explore combinatorial testing for assessing relative completeness in intersectional fairness. By bounding the worst case, we develop a prompt injection scheme within an agent-based framework to enforce conditional fairness with minimal intervention, validated on state-of-the-art GenAI systems.