Bias in Career Path Prediction: Sources, Impacts, Measurement, Mitigation and role of XAI

Abstract

The rapid integration of artificial intelligence (AI) in various sectors, particularly in career path prediction, has sparked significant concerns about the inherent biases and fairness of these AI systems. The potential for discriminatory outcomes is acute, as these biases can exacerbate existing inequalities in career advancement opportunities. This review paper provides a comprehensive examination of the sources, impacts, measurement, and mitigation of bias within AI systems used for career path prediction and the role of Explainable AI (XAI).

We delve into various sources of bias, including those stemming from data collection, algorithm design, and human decision-making processes, with a focus on the challenges posed by AI in reinforcing stereotypes related to professional capabilities and career potentials. These biases are particularly problematic in AI-driven career recommendation systems, where they can misdirect or limit career opportunities for underrepresented groups. The paper discusses extensive strategies for mitigating bias in AI, encompassing ethical considerations necessary for effective implementation and highlighting the importance of interdisciplinary efforts.

This review addresses the impacts of AI bias on individuals' career prospects and the broader professional landscape, examining current methods to counteract these biases. This includes improved data handling, informed model selection, and rigorous adjustments after model development. We emphasize the need for a comprehensive strategy to combat AI bias in career path prediction, which involves using diverse, representative datasets; enhancing transparency and accountability in AI development; and considering alternative AI models that prioritize fairness and ethical implications. This survey aims to contribute to the ongoing dialogue on creating equitable and unbiased AI systems in career path prediction by detailing robust sources, effects, and strategies for addressing AI bias.