The practical ethics of bias reduction in machine translation: why domain adaptation is better than data debiasing

By Marcus Tomalin, Bill Byrne, Shauna Julia Concannon, Danielle Saunders, Stefanie Ullmann. 2021

46 citations recorded by the source corpus.

This article probes the practical ethical implications of AI system design by reconsidering the important topic of bias in the datasets used to train autonomous intelligent systems. The discussion draws on recent work concerning behaviour-guiding technologies, and it adopts a cautious form of technological utopianism by assuming it is potentially beneficial for society at large if AI systems are designed to be comparatively free from the biases that characterise human behaviour. However, the argument presented here critiques the common well-intentioned requirement that, in order to achieve this, all such datasets must be debiased prior to training. By focusing specifically on gender-bias in Neural Machine Translation (NMT) systems, three automated strategies for the removal of bias are considered – downsampling, upsampling, and counterfactual augmentation – and it is shown that systems trained on datasets debiased using these approaches all achieve general translation performance that is much worse than a baseline system. In addition, most of them also achieve worse performance in relation to metrics that quantify the degree of gender bias in the system outputs. By contrast, it is shown that the technique of domain adaptation can be effectively deployed to debias existing NMT systems after they have been fully trained. This enables them to produce translations that are quantitatively far less biased when analysed using gender-based metrics, but which also achieve state-of-the-art general performance. It is hoped that the discussion presented here will reinvigorate ongoing debates about how and why bias can be most effectively reduced in state-of-the-art AI systems.

Subfields: Artificial Intelligence, Information Systems, Natural Language Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Information Systems

In Natural Language Processing

Citations in the corpus

Societal Biases in Language Generation: Progress and Challenges (2021, background): “…(Escud´e Font and Costa-juss`a, 2019; Stanovsky et al., 2019; Saunders and Byrne, 2020; Saunders et al., 2020; Costa-juss`a and de Jorge, 2020; Basta et al., 2020; Stafanoviˇcs et al., 2020; Renduchintala and Williams, 2021; Choubey et al., 2021; Saunders et al., 2021; Tomalin et al., 2021).”
Improving Gender Translation Accuracy with Filtered Self-Training (2021, unknown): “Similar approaches are taken in (Costa-jussà and de Jorge, 2020; Tomalin et al., 2021).”
Limits and Possibilities for “Ethical AI” in Open Source: A Study of Deepfakes (2022, methodology): “We define Implementation harms as those arising through code, algorithm or data problems that can be fixed without changing the intent, or use, of the software, for example through the use of “de-biasing” techniques to reduce bias in algorithms or training data [9, 18, 78].”
Biases in Large Language Models: Origins, Inventory, and Discussion (2023, background): “For example, domain adaptation aims at ine-tuning an existing model with a considerably smaller amount of balanced, ideally unbiased, data [119].”