Robust Bias Detection in MLMs and its Application to Human Trait Ratings
This work addresses the problem of robust bias detection in MLMs for researchers and practitioners, though it is incremental as it builds on prior template-based methods with statistical improvements.
The authors tackled limitations in existing methods for detecting bias in masked language models (MLMs) by proposing a systematic statistical approach using mixed models and pseudo-perplexity weights, which replicated prior studies with small to medium effect sizes. They then applied this method to analyze gender bias in personality and character traits across seven MLMs, finding varying bias patterns such as ALBERT being unbiased for binary gender but most biased for non-binary gender, and some alignment with human psychology findings.
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.