Gender Slopes

Joo, Jungseock; Kärkkäinen, Kimmo

doi:10.1145/3422841.3423533

Cited by 24 publications

(3 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Such an approach has yet to be explored for facial expression analysis. In the domain of facial analysis, counterfactuals have been used to identify [13,27] and mitigate for bias [10]. Our research resembles that of [13] and [27] in that we used a generative adversarial network (GAN), STGAN [33], to generate adversarial counterfactual facial images to assess for counterfactual fairness.…”

Section: Counterfactuals and Biasmentioning

confidence: 99%

“…Though alike in spirit, our paper differs as follow. First, the above studies focused on investigating different methods for counterfactual generation [10,13,27]. In our case, we do not propose an alternative method to generate adversarial or counterfactual images.…”

Section: Counterfactuals and Biasmentioning

confidence: 99%

“…Second, [13] and [27] focused on using the generated counterfactual images to measure the bias present in either a publicly available dataset [13] or existing black-box image analysis APIs [27] but did not propose any method to mitigate bias. Though we do conduct a bias analysis of the model's performance on counterfactuals, the focus is chiefly on deploying counterfactuals explicitly to mitigate for bias which was attempted by [10] as well.…”

Section: Counterfactuals and Biasmentioning

confidence: 99%

See 2 more Smart Citations

Counterfactual Fairness for Facial Expression Recognition

Cheong

Kalkan

Güneş

2023

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Given the increasing prevalence of facial analysis technology, the problem of bias in these tools is becoming an even greater source of concern. Causality has been proposed as a method to address the problem of bias, giving rise to the popularity of using counterfactuals as a bias mitigation tool. In this paper, we undertake a systematic investigation of the usage of counterfactuals to achieve both statistical and causal-based fairness in facial expression recognition. We explore bias mitigation strategies with counterfactual data augmentation at the preprocessing, in-processing, and post-processing stages as well as a stacked approach that combines all three methods. At the in-processing stage, we propose using Siamese Networks to suppress the differences between the predictions on the original and the counterfactual images. Our experimental results on RAF-DB with counterfactuals added show that: (1) The in-processing method outperforms at the pre-processing and postprocessing stages, in terms of accuracy, F1 score, statistical fairness and counterfactual fairness, and (2) stacking the pre-processing, in-processing and post-processing stages provides the best performance.

show abstract