Mitigating Gender Bias for Neural Dialogue Generation with Adversarial Learning

TL;DR

This work tackles gender bias in neural dialogue by introducing Debiased-Chat, an adversarial framework that disentangles unbiased gender features $\oldsymbol{f^{(u)}}$ from biased content via a dedicated disentanglement model and an adversarial dialogue generator. The approach uses an unbiased gendered utterance corpus to train the disentanglement module and applies multiple discriminators to ensure biased features are excluded while preserving meaningful gender cues through unbiased features. Empirical results on the Twitter and Reddit dialogue datasets show substantial bias reduction across offense, sentiment, and gender-word usage metrics, with maintained relevance and higher diversity compared to baselines like CDA and WER. The framework is adaptable to new bias definitions by updating the unbiased utterance corpus, and the authors release their implementation for public use, enabling broader impact in fairer conversational AI.

Abstract

Dialogue systems play an increasingly important role in various aspects of our daily life. It is evident from recent research that dialogue systems trained on human conversation data are biased. In particular, they can produce responses that reflect people's gender prejudice. Many debiasing methods have been developed for various NLP tasks, such as word embedding. However, they are not directly applicable to dialogue systems because they are likely to force dialogue models to generate similar responses for different genders. This greatly degrades the diversity of the generated responses and immensely hurts the performance of the dialogue models. In this paper, we propose a novel adversarial learning framework Debiased-Chat to train dialogue models free from gender bias while keeping their performance. Extensive experiments on two real-world conversation datasets show that our framework significantly reduces gender bias in dialogue models while maintaining the response quality. The implementation of the proposed framework is released.

Figures (2)

• Figure 1: An overview of our proposed framework. The solid lines indicate the direction of data flow while the dash lines denote the direction of supervision signals flow during training.
• Figure 2: A visualization of the disentangled features using t-SNE plot. Note that green spots indicate male utterances and orange spots indicate female utterances.