The Hrunting of AI: Where and How to Improve English Dialectal Fairness

TL;DR

This paper investigates the challenges of improving large language models' performance on English dialects with low data availability, highlighting the impact of human agreement patterns and proposing careful data evaluation for fairness.

Contribution

It reveals how human agreement influences LLM evaluation and discusses the implications for dialectal fairness, emphasizing the need for new tools in low-resource settings.

Findings

LLM-human agreement affects LLM evaluation accuracy

Fine-tuning may not improve and can worsen dialectal performance

Some LLMs can generate high-quality data for scarce dialects

Abstract

It is known that large language models (LLMs) underperform in English dialects, and that improving them is difficult due to data scarcity. In this work we investigate how quality and availability impact the feasibility of improving LLMs in this context. For this, we evaluate three rarely-studied English dialects (Yorkshire, Geordie, and Cornish), plus African-American Vernacular English, and West Frisian as control. We find that human-human agreement when determining LLM generation quality directly impacts LLM-as-a-judge performance. That is, LLM-human agreement mimics the human-human agreement pattern, and so do metrics such as accuracy. It is an issue because LLM-human agreement measures an LLM's alignment with the human consensus; and hence raises questions about the feasibility of improving LLM performance in locales where low populations induce low agreement. We also note that fine-tuning does not eradicate, and might amplify, this pattern in English dialects. But also find encouraging signals, such as some LLMs' ability to generate high-quality data, thus enabling scalability. We argue that data must be carefully evaluated to ensure fair and inclusive LLM improvement; and, in the presence of scarcity, new tools are needed to handle the pattern found.