Robust Bias Detection in MLMs and its Application to Human Trait Ratings

TL;DR

This paper introduces a systematic statistical method to measure bias in masked language models, accounting for variability and effect sizes, and applies it to gender bias in personality and character traits across multiple models.

Contribution

It develops a novel bias quantification approach using mixed models and applies it to analyze gender bias in MLMs for personality and character traits, revealing model-specific biases and some alignment with human psychology.

Findings

ALBERT is unbiased for binary gender but biased for non-binary neo.

RoBERTa-large shows bias for binary gender but less for neo.

Some alignment exists between MLM bias and human psychology findings.

Abstract

There has been significant prior work using templates to study bias against demographic attributes in MLMs. However, these have limitations: they overlook random variability of templates and target concepts analyzed, assume equality amongst templates, and overlook bias quantification. Addressing these, we propose a systematic statistical approach to assess bias in MLMs, using mixed models to account for random effects, pseudo-perplexity weights for sentences derived from templates and quantify bias using statistical effect sizes. Replicating prior studies, we match on bias scores in magnitude and direction with small to medium effect sizes. Next, we explore the novel problem of gender bias in the context of $\textit{personality}$ and $\textit{character}$ traits, across seven MLMs (base and large). We find that MLMs vary; ALBERT is unbiased for binary gender but the most biased for non-binary $\textit{neo}$, while RoBERTa-large is the most biased for binary gender but shows small to no bias for $\textit{neo}$. There is some alignment of MLM bias and findings in psychology (human perspective) - in $\textit{agreeableness}$ with RoBERTa-large and $\textit{emotional stability}$ with BERT-large. There is general agreement for the remaining 3 personality dimensions: both sides observe at most small differences across gender. For character traits, human studies on gender bias are limited thus comparisons are not feasible.