Where Do Self-Supervised Speech Models Become Unfair?

TL;DR

This study investigates how self-supervised speech models exhibit layerwise biases against certain speaker groups, revealing biases are established early and persist through fine-tuning.

Contribution

First layerwise fairness analysis of pretrained self-supervised speech models, showing biases are present from initial layers and are difficult to eliminate after fine-tuning.

Findings

Bias against certain speaker groups appears from the first layers.

Opposite layerwise bias patterns are observed for speaker identification and speech recognition.

Biases established during pretraining are hard to remove after fine-tuning.

Abstract

Speech encoder models are known to model members of some speaker groups (SGs) better than others. However, there has been little work in establishing why this occurs on a technological level. To our knowledge, we present the first layerwise fairness analysis of pretrained self-supervised speech encoder models (S3Ms), probing each embedding layer for speaker identification (SID) automatic speech recognition (ASR). We find S3Ms produce embeddings biased against certain SGs for both tasks, starting at the very first latent layers. Furthermore, we find opposite patterns of layerwise bias for SID vs ASR for all models in our study: SID bias is minimized in layers that minimize overall SID error; on the other hand, ASR bias is maximized in layers that minimize overall ASR error. The inverse bias/error relationship for ASR is unaffected when probing S3Ms that are finetuned for ASR, suggesting SG-level bias is established during pretraining and is difficult to remove.