Wisdom from Diversity: Bias Mitigation Through Hybrid Human-LLM Crowds

TL;DR

This paper investigates bias mitigation in large language models by analyzing responses, demonstrating that hybrid human-LLM crowds with weighted aggregation strategies can effectively reduce biases and improve accuracy.

Contribution

It introduces hybrid human-LLM crowds and locally weighted aggregation methods as novel strategies for bias mitigation and performance enhancement in LLMs.

Findings

Averaging LLM responses can increase bias due to limited diversity.

Locally weighted aggregation reduces bias and improves accuracy.

Hybrid crowds of humans and LLMs further decrease biases and enhance performance.

Abstract

Despite their performance, large language models (LLMs) can inadvertently perpetuate biases found in the data they are trained on. By analyzing LLM responses to bias-eliciting headlines, we find that these models often mirror human biases. To address this, we explore crowd-based strategies for mitigating bias through response aggregation. We first demonstrate that simply averaging responses from multiple LLMs, intended to leverage the "wisdom of the crowd", can exacerbate existing biases due to the limited diversity within LLM crowds. In contrast, we show that locally weighted aggregation methods more effectively leverage the wisdom of the LLM crowd, achieving both bias mitigation and improved accuracy. Finally, recognizing the complementary strengths of LLMs (accuracy) and humans (diversity), we demonstrate that hybrid crowds containing both significantly enhance performance and further reduce biases across ethnic and gender-related contexts.