Confronting LLMs with Traditional ML: Rethinking the Fairness of Large Language Models in Tabular Classifications

TL;DR

This paper investigates how large language models (LLMs) inherit social biases from their training data, affecting fairness in tabular classification tasks, and compares their bias mitigation effectiveness to traditional models.

Contribution

It reveals that LLMs inherently carry social biases from pretraining data and that bias mitigation techniques have limited success compared to traditional models.

Findings

LLMs inherit social biases from training data.

Bias mitigation methods only moderately reduce biases.

Bias gap remains larger in LLMs than in traditional models.

Abstract

Recent literature has suggested the potential of using large language models (LLMs) to make classifications for tabular tasks. However, LLMs have been shown to exhibit harmful social biases that reflect the stereotypes and inequalities present in society. To this end, as well as the widespread use of tabular data in many high-stake applications, it is important to explore the following questions: what sources of information do LLMs draw upon when making classifications for tabular tasks; whether and to what extent are LLM classifications for tabular data influenced by social biases and stereotypes; and what are the consequential implications for fairness? Through a series of experiments, we delve into these questions and show that LLMs tend to inherit social biases from their training data which significantly impact their fairness in tabular classification tasks. Furthermore, our investigations show that in the context of bias mitigation, though in-context learning and finetuning have a moderate effect, the fairness metric gap between different subgroups is still larger than that in traditional machine learning models, such as Random Forest and shallow Neural Networks. This observation emphasizes that the social biases are inherent within the LLMs themselves and inherited from their pretraining corpus, not only from the downstream task datasets. Besides, we demonstrate that label-flipping of in-context examples can significantly reduce biases, further highlighting the presence of inherent bias within LLMs.