Gender bias in Artificial Intelligence: ARTICLES

This report examines the opportunities and challenges for gender equality in labour markets transformed by artificial intelligence (AI) and platform work.

Digital transformation and artificial intelligence (AI) are an integral part of the 4th Industrial Revolution, transforming our jobs and lives. But, while AI could well be a key driver for innovation, there are challenges and risks that cannot be ignored. This is why the Commission’s White Paper on AI stresses the importance of building an AI ‘ecosystem of trust’ with ‘rules that put people at the centre’, as President von der Leyen stated in her 2020 State of the Union address.

Cet article propose une réflexion sur les pratiques d’évaluation des systèmes de reconnaissance automatique de la parole (ASR) et s'intéresse aux pratiques actuelles lors des grandes campagnes d’évaluation. Il observe que la variabilité de la parole et plus particulièrement celle de l’individu n’est pas prise en compte dans les protocoles d’évaluation actuels rendant impossible l’étude de biais potentiels dans les systèmes.

This article explores how biased data inconveniences and even endangers women's lives and sketches possible solutions to this problem.

Women’s health issues are routinely underreported, underdiagnosed, and undertreated.10 This Paper highlights a few considerations related to digital health innovation, particularly artificial intelligence and machine learning, and women’s health.

This article takes a look at the way gender bias is created and shows four way to minimise it.

Organizations need to pay more attention to the diversity of the teams developing their AI solutions. Doing so will help prevent gender bias and maximise AI’s potential to transform the workplace.

Amazon.com Inc's AMZN.O machine-learning specialists uncovered a big problem: their new recruiting engine did not like women.

Computer scientists must identify sources of bias, de-bias training data and develop artificial-intelligence algorithms that are robust to skews in the data, argue James Zou and Londa Schiebinger.

Recent studies demonstrate that machine learning algorithms can discriminate based on classes like race and gender. This work presents an approach to evaluate bias present in automated facial analysis algorithms and datasets with respect to phenotypic subgroups.

This paper presents research uncovering systematic gender bias in the representation of political leaders in the media, using artificial intelligence. Newspaper coverage of Irish ministers over a fifteen year period was gathered and analysed with AI. Findings demonstrate evidence of gender bias in regard to female politicians. This paper also sets out a methodology whereby media content may be analysed on a large scale.

This paper draws on a substantial body of theoretical and empirical research concerning how digital disruption and artificial intelligence alter industry structures and sex-based divisions of labor, and using and replicating data from various sources performed analyses and made estimates regarding patterns of jobs lost and gained (% of 2019 employment for each gender) and potential impact of job automation over time across workers (%, by gender, age group, and education level).

Abusive language detection models tend to have a problem of being biased toward identity words of a certain group of people because of imbalanced training datasets. Such model bias is an obstacle for models to be robust enough for practical use. This work measures gender biases on models trained with different abusive language datasets, while analyzing the effect of different pre-trained word embeddings and model architecturesand also experiment with three different bias mitigation methods.

This thesis (1) characterizes the gender and skin type distribution of IJB-A, a government facial recognition benchmark, and Adience, a gender classification benchmark, (2) outlines an approach for capturing images with more diverse skin types which is then applied to develop the Pilot Parliaments Benchmark (PPB), and (3) uses PPB to assess the classification accuracy of Adience, IBM, Microsoft, and Face++ gender classifiers with respect to gender, skin type, and the intersection of skin type and gender.