"Thy algorithm shalt not bear false witness": An Evaluation of Multiclass Debiasing Methods on Word Embeddings

Thalea Schlender, Gerasimos Spanakis

Research output: Chapter in Book/Report/Conference proceedingConference article in proceedingAcademicpeer-review

Abstract

With the vast development and employment of artificial intelligence applications, research into the fairness of these algorithms has been increased. Specifically, in the natural language processing domain, it has been shown that social biases persist in word embeddings and are thus in danger of amplifying these biases when used. As an example of social bias, religious biases are shown to persist in word embeddings and the need for its removal is highlighted. This paper investigates the state-of-the-art multiclass debiasing techniques: Hard debiasing, SoftWEAT debiasing and Conceptor debiasing. It evaluates their performance when removing religious bias on a common basis by quantifying bias removal via the Word Embedding Association Test (WEAT), Mean Average Cosine Similarity (MAC) and the Relative Negative Sentiment Bias (RNSB). By investigating the religious bias removal on three widely used word embeddings, namely: Word2Vec, GloVe, and ConceptNet, it is shown that the preferred method is ConceptorDebiasing. Specifically, this technique manages to decrease the measured religious bias on average by 82,42%, 96,78% and 54,76% for the three word embedding sets respectively.

Original languageEnglish
Title of host publicationBNAIC/BeneLearn 2020
EditorsLu Cao, Walter Kosters, Jefrey Lijffijt
Pages254-268
Number of pages15
Publication statusPublished - Nov 2020
EventBenelux Conference on Artificial Intelligence and Machine Learning - Online, Leiden University, Leiden, Netherlands
Duration: 19 Nov 202020 Nov 2020
https://bnaic.liacs.leidenuniv.nl/

Conference

ConferenceBenelux Conference on Artificial Intelligence and Machine Learning
Abbreviated titleBNAIC/BeneLearn 2020
Country/TerritoryNetherlands
CityLeiden
Period19/11/2020/11/20
Internet address

Cite this