Thinking Fair and Slow: On the Efficacy of Structured Prompts for Debiasing Language Models

TL;DR

This paper explores the use of structured, System 2-inspired prompts to reduce bias in language model outputs, offering an accessible alternative to traditional debiasing methods.

Contribution

It introduces a novel iterative prompting framework that applies logical and reflective prompts to improve fairness in LLM outputs, demonstrating effectiveness across multiple models and datasets.

Findings

Structured prompts significantly reduce bias in LLM outputs.

System 2-based implicative prompts outperform other prompting strategies.

The approach maintains competitive performance on downstream tasks.

Abstract

Existing debiasing techniques are typically training-based or require access to the model's internals and output distributions, so they are inaccessible to end-users looking to adapt LLM outputs for their particular needs. In this study, we examine whether structured prompting techniques can offer opportunities for fair text generation. We evaluate a comprehensive end-user-focused iterative framework of debiasing that applies System 2 thinking processes for prompts to induce logical, reflective, and critical text generation, with single, multi-step, instruction, and role-based variants. By systematically evaluating many LLMs across many datasets and different prompting strategies, we show that the more complex System 2-based Implicative Prompts significantly improve over other techniques demonstrating lower mean bias in the outputs with competitive performance on the downstream tasks. Our work offers research directions for the design and the potential of end-user-focused evaluative frameworks for LLM use.