In today’s world, artificial intelligence has become ubiquitous, touching almost every facet of our lives. It operates on algorithms, which are built upon data. Yet, when this data contains biases and prejudices against specific groups, there’s a significant risk of them being perpetuated. From everyday problems like automatic soap dispensers that fail to recognize certain skin tones to more consequential disparities in job application assessments or court proceedings, these biases manifest in various forms.

We aim to understand the formation and evolution of inequalities and biases that are caused by the connectivity patterns between people and entities within a complex system. Those inequalities and biases include algorithmic unfairness, inequality of visibility, perception biases, structural marginality, and persistence of inequalities over time. As we live in complex socio-economic networks, the structure of these relationships not only limits one’s social view, behavioral decisions, and social influence but also affects AI systems and algorithms. We employ methods borrowed from data science, mathematics, machine learning, and statistical physics.

We aim to find effective methods to tackle structural inequalities and reduce the effects of their consequences. We search for causal dynamics, and based on it, design AI systems that are explainable, responsible, interpretable, and resilient. Ultimately, reducing inequalities benefits all members of society and will help create a better future for the next generation.