Effectiveness of Debiasing Techniques: An Indigenous Qualitative Analysis
This work tackles the issue of algorithmic bias for underrepresented indigenous communities, offering a domain-specific approach that is incremental in adapting existing methods to new cultural contexts.
The paper addresses the problem that existing debiasing techniques for pre-trained language models are biased towards US racial contexts and fail to represent indigenous populations like Māori in New Zealand, proposing an indigenous qualitative analysis to incorporate local knowledge for more effective debiasing.
An indigenous perspective on the effectiveness of debiasing techniques for pre-trained language models (PLMs) is presented in this paper. The current techniques used to measure and debias PLMs are skewed towards the US racial biases and rely on pre-defined bias attributes (e.g. "black" vs "white"). Some require large datasets and further pre-training. Such techniques are not designed to capture the underrepresented indigenous populations in other countries, such as Māori in New Zealand. Local knowledge and understanding must be incorporated to ensure unbiased algorithms, especially when addressing a resource-restricted society.