From Algorithm Design to Algorithmic Bias: Identifying Discriminatory Risks in Innovation Management

The growing integration of Generative Artificial Intelligence into managerial processes has enhanced decision-making, data interpretation, and operational efficiency. However, alongside these benefits, algorithmic bias has emerged as a critical concern, particularly when AI systems inadvertently reinforce existing social inequalities. This study investigates how such biases are introduced during the algorithm design, a phase where technical and personal assumptions become embedded in AI systems. Through a systematic literature review, four key areas of bias emergence are identified: gendered anthropomorphism, proxy variable selection, dataset composition and labeling, and lack of human oversight in algorithm selection. The findings emphasize that mitigating algorithmic bias requires inclusive design, diverse teams, and strong oversight, aligning AI systems development with ethical principles and regulatory standards.