English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

To mitigate gender bias in contextualized language models, differentintrinsic mitigation strategies have been proposed, alongside many biasmetrics. Considering that the end use of these language models is fordownstream tasks like text classification, it is important to understand howthese intrinsic bias mitigation strategies actually translate to fairness indownstream tasks and the extent of this. In this work, we design a probe toinvestigate the effects that some of the major intrinsic gender bias mitigationstrategies have on downstream text classification tasks. We discover thatinstead of resolving gender bias, intrinsic mitigation techniques and metricsare able to hide it in such a way that significant gender information isretained in the embeddings. Furthermore, we show that each mitigation techniqueis able to hide the bias from some of the intrinsic bias measures but not all,and each intrinsic bias measure can be fooled by some mitigation techniques,but not all. We confirm experimentally, that none of the intrinsic mitigationtechniques used without any other fairness intervention is able to consistentlyimpact extrinsic bias. We recommend that intrinsic bias mitigation techniquesshould be combined with other fairness interventions for downstream tasks.

How Far Can It Go?: On Intrinsic Gender Bias Mitigation for Text Classification