What Triggers my Model? Contrastive Explanations Inform Gender Choices by Translation Models

TL;DR

This paper investigates the origins of gender bias in neural machine translation by analyzing which source tokens influence gender choices, using contrastive explanations and saliency attribution to compare model and human perceptions.

Contribution

It introduces a contrastive attribution method to identify source tokens influencing gender decisions in translation models, linking model behavior to human perceptions.

Findings

Salient words identified overlap with human perceptions

Contrastive attribution reveals key context influencing gender choices

Analysis highlights potential for bias mitigation strategies

Abstract

Interpretability can be implemented to understand decisions taken by (black box) models, such as neural machine translation (NMT) or large language models (LLMs). Yet, research in this area has been limited in relation to a manifested problem in these models: gender bias. In this work, we aim to move away from simply measuring bias to exploring its origins. Working with gender-ambiguous natural source data, this exploratory study examines which context, in the form of input tokens in the source sentence (EN), influences (or triggers) the NMT model's choice of a certain gender inflection in the target languages (DE/ES). To analyse this, we compute saliency attribution based on contrastive translations. We first address the challenge of the lack of a scoring threshold and specifically examine different attribution levels of source words on the model's gender decisions in the translation. We compare salient source words with human perceptions of gender and demonstrate a noticeable overlap between human perceptions and model attribution. Additionally, we provide a linguistic analysis of salient words. Our work showcases the relevance of understanding model translation decisions in terms of gender, how this compares to human decisions and that this information should be leveraged to mitigate gender bias.