Applying Intrinsic Debiasing on Downstream Tasks: Challenges and Considerations for Machine Translation
It addresses the gap between intrinsic bias removal and practical applications in machine translation, highlighting incremental improvements for bias mitigation in NLP.
The paper investigates how intrinsic debiasing methods affect neural machine translation, identifying three key challenges that impact downstream performance and debiasing success.
Most works on gender bias focus on intrinsic bias -- removing traces of information about a protected group from the model's internal representation. However, these works are often disconnected from the impact of such debiasing on downstream applications, which is the main motivation for debiasing in the first place. In this work, we systematically test how methods for intrinsic debiasing affect neural machine translation models, by measuring the extrinsic bias of such systems under different design choices. We highlight three challenges and mismatches between the debiasing techniques and their end-goal usage, including the choice of embeddings to debias, the mismatch between words and sub-word tokens debiasing, and the effect on different target languages. We find that these considerations have a significant impact on downstream performance and the success of debiasing.