Wisdom from Diversity: Bias Mitigation Through Hybrid Human-LLM Crowds
This addresses bias mitigation in LLMs for applications requiring fair and accurate outputs, representing an incremental improvement through hybrid aggregation methods.
The paper tackles bias in large language models by showing that simple averaging of multiple LLM responses exacerbates biases, while locally weighted aggregation methods achieve better bias mitigation and accuracy, and hybrid human-LLM crowds further reduce biases and enhance performance across ethnic and gender contexts.
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.