The Role of Biased Data in Computerized Gender Discrimination

Abstract

Gender bias is prevalent in all walks of life from schools to colleges, corporate as well as government offices. This has led to the under-representation of the female gender in many professions. Most of the Artificial Intelligence-Natural Language Processing (AI-NLP) models learning from these underrepresented real world datasets amplify the bias in many cases, resulting in traditional biases being reinforced. In this paper, we have discussed how gender bias became ingrained in our society and how it results in the underrepresentation of the female gender in several fields such as education, healthcare, STEM, film industry, food industry, and sports. We shed some light on how traditional gender bias is reflected in AI-NLP systems such as automated resume screening, machine translation, text generation, etc. Future prospects of these AI-NLP applications need to include possible solutions to these existing biased AI-NLP applications, such as debiasing the word embeddings and having guidelines for more ethical and transparent standards. ACM Reference Format: Md. Arshad Ahmed, Madhura Chatterjee, Pankaj Dadure, and Partha Pakray. 2022. The Role of Biased Data in Computerized Gender Discrimination. In Third Workshop on Gender Equality, Diversity, and Inclusion in Software Engineering (GE@ICSE’22), May 20, 2022, Pittsburgh, PA, USA. ACM, New York, NY, USA, 6 pages.