Responsible AI: Gender bias assessment in emotion recognition

Domnich, Artem; Anbarjafari, Gholamreza

doi:10.48550/arxiv.2103.11436

Cited by 9 publications

(11 citation statements)

References 84 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…With the advancement of AI, many are concerned there is a high chance that AI development can lead to significant and irreversible repercussions since there are substantial developments in affective computing, considerable amounts of work concerning facial expression and deep learning. In [19], the authors provided an overview of FER biases in the context of responsible AI. They focused on gender bias and stated that race is another bias that needs to be addressed.…”

Section: Related Workmentioning

confidence: 99%

Ethical AI in facial expression analysis: racial bias

Sham

Aktas

Rizhinashvili

et al. 2022

SIViP

Self Cite

View full text Add to dashboard Cite

Facial expression recognition using deep neural networks has become very popular due to their successful performances. However, the datasets used during the development and testing of these methods lack a balanced distribution of races among the sample images. This leaves a possibility of the methods being biased toward certain races. Therefore, a concern about fairness arises, and the lack of research aimed at investigating racial bias only increases the concern. On the other hand, such bias in the method would decrease the real-world performance due to the wrong generalization. For these reasons, in this study, we investigated the racial bias within popular state-of-the-art facial expression recognition methods such as Deep Emotion, Self-Cure Network, ResNet50, InceptionV3, and DenseNet121. We compiled an elaborated dataset with images of different races, cross-checked the bias for methods trained, and tested on images of people of other races. We observed that the methods are inclined towards the races included in the training data. Moreover, an increase in the performance increases the bias as well if the training dataset is imbalanced. Some methods can make up for the bias if enough variance is provided in the training set. However, this does not mitigate the bias completely. Our findings suggest that an unbiased performance can be obtained by adding the missing races into the training data equally. KeywordsFacial expression recognition (FER) • Deep neural networks • Reaction emotion • LSTM Abdallah Hussein Sham and Kadir Aktas are both equally led this work.Our thanks to Pexels API for granting us the rights for the data collection of the database.

show abstract

Section: Related Workmentioning

confidence: 99%

Ethical AI in facial expression analysis: racial bias

Sham

Aktas

Rizhinashvili

et al. 2022

SIViP

Self Cite

View full text Add to dashboard Cite

show abstract

“…These subdomains can include different aspects of possible bias sources, be it gender, race, age, and others. For example, in [2], authors show which models are gender-biased, which are not, and how the gender of the subject affects its emotion recognition. They also describe the extent of this bias by measuring the accuracy gap in emotion recognition between male and female test sets and observing which types of emotions are better classified for men and women.…”

Section: Related Workmentioning

confidence: 99%

“…In recent times, research in artificial intelligence (AI) and machine learning (ML) techniques have led to significant improvements in computer vision, speech processing, and language technologies, among others. Consequently, with these advances has come an inadvertent focus on the ethics of such ML models [1][2][3].…”

Section: Introductionmentioning

confidence: 99%

Gender Neutralisation for Unbiased Speech Synthesising

Rizhinashvili

Sham²,

Anbarjafari

2022

Electronics

Self Cite

View full text Add to dashboard Cite

Machine learning can encode and amplify negative biases or stereotypes already present in humans, resulting in high-profile cases. There can be multiple sources encoding the negative bias in these algorithms, like errors from human labelling, inaccurate representation of different population groups in training datasets, and chosen model structures and optimization methods. Our paper proposes a novel approach to speech processing that can resolve the gender bias problem by eliminating the gender parameter. Therefore, we devised a system that transforms the input sound (speech of a person) into a neutralized voice to the point where the gender of the speaker becomes indistinguishable by both humans and AI. Wav2Vec based network has been utilised to conduct speech gender recognition to validate the main claim of this research work, which is the neutralisation of gender from the speech. Such a system can be used as a batch pre-processing layer for training models, thus making associated gender bias irrelevant. Further, such a system can also find its application where speaker gender bias by humans is also prominent, as the listener will not be able to judge the gender from speech.

show abstract

“…Social biases in downstream tasks expose users with multiple disadvantaged sensitive attributes to unknown but potentially harmful outcomes, especially when models trained on downstream tasks are used in real-world decision making, such as for screening résumes or predicting recidivism in criminal proceedings (Bolukbasi et al, 2016;Angwin et al, 1999). In this work, we choose emotion regression as a downstream task because social biases are often realized through emotion recognition (Elfenbein and Ambady, 2002) and machine learning models have been shown to reflect gender bias in emotion recognition tasks (Domnich and Anbarjafari, 2021). For example, sentiment analysis and emotion regression may be used by companies to measure product engagement for different social groups.…”

Section: Introductionmentioning

confidence: 99%

Mapping the Multilingual Margins: Intersectional Biases of Sentiment Analysis Systems in English, Spanish, and Arabic

Câmara¹,

Taneja²,

Azad³

et al. 2022

Proceedings of the Second Workshop on Language Technology for Equality, Diversity and Inclusion

View full text Add to dashboard Cite

As natural language processing systems become more widespread, it is necessary to address fairness issues in their implementation and deployment to ensure that their negative impacts on society are understood and minimized. However, there is limited work that studies fairness using a multilingual and intersectional framework or on downstream tasks. In this paper, we introduce four multilingual Equity Evaluation Corpora, supplementary test sets designed to measure social biases, and a novel statistical framework for studying unisectional and intersectional social biases in natural language processing. We use these tools to measure gender, racial, ethnic, and intersectional social biases across five models trained on emotion regression tasks in English, Spanish, and Arabic. We find that many systems demonstrate statistically significant unisectional and intersectional social biases. 1

show abstract

Responsible AI: Gender bias assessment in emotion recognition

Cited by 9 publications

References 84 publications

Ethical AI in facial expression analysis: racial bias

Ethical AI in facial expression analysis: racial bias

Gender Neutralisation for Unbiased Speech Synthesising

Mapping the Multilingual Margins: Intersectional Biases of Sentiment Analysis Systems in English, Spanish, and Arabic

Contact Info

Product

Resources

About