Artificial Intelligence Bias and the Amplification of Inequalities in the Labor Market

Artificial intelligence (AI) is now present in nearly every aspect of our daily lives. Furthermore, while this AI augmentation is generally beneficial, or at worst, nonproblematic, some instances warrant attention. In this study, we argue that AI bias resulting from training data sets in the labor market can significantly amplify minor inequalities, which later in life manifest as permanently lost opportunities and social status and wealth segregation. The Matthew effect is responsible for this phenomenon, except that the focus is not on the rich getting richer, but on the poor becoming even poorer. We demonstrate how frequently changing expectations for skills, competencies, and knowledge lead to AI failing to make impartial hiring decisions. Specifically, the bias in the training data sets used by AI affects the results, causing the disadvantaged to be overlooked while the privileged are frequently chosen. This simple AI bias contributes to growing social inequalities by reinforcing the Matthew effect, and it does so at much faster rates than previously. We assess these threats by studying data from various labor fields, including justice, security, healthcare, human resource management, and education.