Causally fair text generation reduces bias in large language models
Causally Fair Generation with Large Language Models
Artificial IntelligenceMachine Learning
Summary
Large language models are tools that generate text and help make decisions, but they can sometimes produce biased or unfair outcomes that affect different groups differently. The authors present a method called CFG that uses ideas from cause-and-effect reasoning to identify and remove unfair biases in the text generated by these models. This method lets users specify which biases to remove while keeping important, fair reasons for certain differences. They tested CFG on multiple models and datasets, showing it can reduce bias while keeping useful information.
What this means in practice
- •For machine learning engineers: Implement a system that generates text outputs free from selected unfair biases for use in AI-powered decision processes.
- •For social science data teams: Analyze text generation effects on demographic groups by controlling causal pathways that affect fairness outcomes in social datasets.
Authors
Patrik Okanovic, Torsten Hoefler, Drago Plecko
Abstract
Large language models (LLMs) are increasingly used to generate, complete, and transform information in settings where their outputs can shape consequential decisions, raising concerns about their impact on demographic disparities. In this context, causal inference provides a principled basis for assessing fairness, because it attributes observed disparities to the mechanisms that generated them, which a purely statistical approach cannot do even with infinite data. In LLM generation, a query may request several causally related variables, each of which is both an outcome of interest and a possible cause of other outputs, and the information supplied in the prompt need not follow a topological or a temporal order. This calls for methods that can analyze and selectively remove disparities from such a flexible generation process. In this paper we introduce Causally Fair Generation with LLMs (CFG, for short). CFG extracts relevant concepts, grounds generation in a reference population and causal diagram, and removes user-selected causal effects. CFG also allows pathways deemed justifiable for the task's utility to be retained, which is known in legal literature as business necessity. Further, we provide formal guarantees for our method when eliminating all discriminatory causal effects in the adapted population model, under appropriate causal assumptions. We evaluate CFG with four LLMs in three real-world settings based on population data and on a synthetic dataset with a known causal ground truth.