Strategic Demonstration Selection for Improved Fairness in LLM In-Context Learning

TL;DR

This paper explores how demonstration selection in in-context learning influences fairness in large language models processing tabular data, proposing a clustering-based mitigation method to improve fairness and performance.

Contribution

It introduces a novel clustering and evolutionary strategy to curate diverse demonstrations, significantly enhancing fairness in LLM in-context learning without compromising accuracy.

Findings

Including minority samples in prompts boosts fairness.

The ratio of minority to majority samples affects fairness-accuracy trade-off.

Proposed clustering method improves fairness metrics substantially.

Abstract

Recent studies highlight the effectiveness of using in-context learning (ICL) to steer large language models (LLMs) in processing tabular data, a challenging task given the structured nature of such data. Despite advancements in performance, the fairness implications of these methods are less understood. This study investigates how varying demonstrations within ICL prompts influence the fairness outcomes of LLMs. Our findings reveal that deliberately including minority group samples in prompts significantly boosts fairness without sacrificing predictive accuracy. Further experiments demonstrate that the proportion of minority to majority samples in demonstrations affects the trade-off between fairness and prediction accuracy. Based on these insights, we introduce a mitigation technique that employs clustering and evolutionary strategies to curate a diverse and representative sample set from the training data. This approach aims to enhance both predictive performance and fairness in ICL applications. Experimental results validate that our proposed method dramatically improves fairness across various metrics, showing its efficacy in real-world scenarios.