The goal of this research is to investigate the detection and mitigation of the gender bias and intersectional bias of Artificial Intelligence (AI) algorithms. More specifically, the aim is to investigate gender bias mitigation in AI algorithms used in recommendation systems and social media with user-centric strategies to understand the societal impact these bias mitigation algorithms have. Previous research shows that bias exists in many AI systems. Recently, efforts have been made to reduce this bias in favour of fair AI that is both inclusive and responsible. However, the focus lies largely on the technical aspect, where debiasing techniques are developed and tested in an offline setting, where accuracy is being measured using different scoring methods. There is a pressing need to evaluate bias detecting and mitigation strategies with user-centred methods. In this research, bias detection algorithms are investigated with an intersectional approach through three cumulative studies. The real-world impact of debiasing techniques used for mitigating gender bias in AI algorithms need to be studied to make AI that actively contribute to an equal and inclusive society. Humans may create AI systems that contribute to a less biased world, but debiased explainable AI systems may also help humans in the process of making a more equal society.

2024