Mitigating Gender Bias in Machine Learning Data Sets

Leavy, Susan; Meaney, Gerardine; Wade, Karen; Greene, Derek

doi:10.1007/978-3-030-52485-2_2

Cited by 39 publications

(18 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Analyses that examine algorithm outcomes in different intersectional subgroups, such as Black and other women of colour (Buolamwini and Gebru, 2018) can help to uncover intersectional impacts. Machine learning techniques can be deployed to uncover and address bias in the training data that feed algorithms (Leavy et al, 2020), and computer scientists are exploring different definitions of 'fairness' in algorithms along with new computational techniques to ameliorate algorithmic bias (Mehrabi et al, 2019). While these efforts will not eradicate technologically-facilitated structural bias, they will reduce its effects as we work towards a multi-faceted approach to TFV more generally.…”

Section: Discussionmentioning

confidence: 99%

Tech-facilitated violence: thinking structurally and intersectionally

Bailey

Burkell

2021

Journal of Gender-Based Violence

View full text Add to dashboard Cite

Technologically-facilitated violence (TFV) can take many shapes and forms, In this thought piece, we reflect on TFV from structural and intersectional perspectives, examining how these might change our understanding of TFV, with particular attention to gender-based TFV. We are motivated to engage in this reflection for two main reasons. First, traditional understandings of violence, including gender-based violence, tend to prioritise physical acts (whether in word or in application), contributing to a trivialisation of the kinds of harms effected through digitised communications networks (Dunn, 2021). Second, if TFV is understood primarily in terms of individual interpersonal acts, our ability to understand how intersecting oppressions such as sexism, racism, homophobia, transphobia, colonialism affect the likelihood of being targeted and the experience of violence will be compromised. As Black feminist and critical race scholars such as Crenshaw (1991), Hill Collins (2017), and Jiwani, Berman and Cameron (2010) have ably demonstrated, individualistic single axis accounts of violence outside of technologised contexts have resulted in exclusionary and dangerous outcomes that selectively harm members of equality-seeking communities. The result of these individualised understandings of violence is that structural oppressions are ‘erased, trivialised, or contained within categories that evacuate the violation of [structural] violence’ (Jiwani, 2006, xi–xii). Among other effects, such erasure risks rendering invisible opportunities to intervene with respect to violence not carried out by individuals, often resulting in ‘remedies’ that emphasise interventions by the state against individual actors (for example, through criminal law), powers that already disproportionately target members of equality-seeking communities, and misses the potential need to intervene on capitalistic corporate systems and behaviours. In both cases, the prospect of achieving justice recedes.<br /><br />Key message<br /><ul><li>Essential to understand TFV through structural and intersectional lenses to better ensure just policy approaches and support mechanisms for all.</li></ul>

show abstract

Section: Discussionmentioning

confidence: 99%

Tech-facilitated violence: thinking structurally and intersectionally

Bailey

Burkell

2021

Journal of Gender-Based Violence

View full text Add to dashboard Cite

show abstract

“…Sheng et al [74] further measure bias associated with sexual orientation by comparing the associated sentiment. We also have the extensive study by Leavy et al [56] on CBOW trained on articles from The Guardian journal and the British Digital Library. In addition, Hutchinson et al [83] studied the perception of models towards disabled people and Bhardwaj et al [47] combined the study of gender bias on BERT by sentiment analysis with gender separability.…”

Section: Association Testsmentioning

confidence: 99%

A Survey on Bias in Deep NLP

Garrido-Muñoz¹,

Montejo-Ráez²,

Martínez-Santiago³

et al. 2021

Applied Sciences

View full text Add to dashboard Cite

Deep neural networks are hegemonic approaches to many machine learning areas, including natural language processing (NLP). Thanks to the availability of large corpora collections and the capability of deep architectures to shape internal language mechanisms in self-supervised learning processes (also known as “pre-training”), versatile and performing models are released continuously for every new network design. These networks, somehow, learn a probability distribution of words and relations across the training collection used, inheriting the potential flaws, inconsistencies and biases contained in such a collection. As pre-trained models have been found to be very useful approaches to transfer learning, dealing with bias has become a relevant issue in this new scenario. We introduce bias in a formal way and explore how it has been treated in several networks, in terms of detection and correction. In addition, available resources are identified and a strategy to deal with bias in deep NLP is proposed.

show abstract

“…However, bias does not arise as a result of algorithm design in most cases; rather, algorithms inherit existing bias from historical data that contains remnants of bias from human decision-making and culture. In this case, the root of this kind of prejudice lies in the way traditional gender ideology and latent discrimination are captured in the data from which the algorithm learns (Leavy et al, 2020). The significant component of data that was utilized for training the AI recruiting system was the resumes of employees in the company, mostly males.…”

Section: Fairness Bias and Discriminationmentioning

confidence: 99%

Machine learning in Governments: Benefits, Challenges and Future Directions

2021

JeDEM

View full text Add to dashboard Cite

The unprecedented increase in computing power and data availability has signifi-cantly altered the way and the scope that organizations make decisions relying on technologies. There is a conspicuous trend that organizations are seeking the use of frontier technologies with the purpose of helping the delivery of services and making day-to-day operational deci-sions. Machine learning (ML) is the fastest growing and at the same time, the most debated and controversial of these technologies. Although there is a great deal of research in the literature related to machine learning applications, most of them focus on the technical aspects or pri-vate sector use. The governmental machine learning applications suffer the lack of theoretical and empirical studies and unclear governance framework. This paper reviews the literature on the use of machine learning by government, aiming to identify the benefits and challenges of wider adoption of machine learning applications in the public sector and to propose the direc-tions for future research.

show abstract

Mitigating Gender Bias in Machine Learning Data Sets

Cited by 39 publications

References 22 publications

Tech-facilitated violence: thinking structurally and intersectionally

Tech-facilitated violence: thinking structurally and intersectionally

A Survey on Bias in Deep NLP

Machine learning in Governments: Benefits, Challenges and Future Directions

Contact Info

Product

Resources

About