From performance to fairness-awareness: Morality in organization's use of AI

Vancompernolle Vromman, Flore [UCL] Vande Kerckhove, Corentin [UCL] Hericher, Corentin [UCL]

The rapid integration of artificial intelligence (AI) and algorithms in organizational decision-making has recently prompted concerns about the ethical implication of it. While algorithms offer performance-related advantages, their outputs may perpetuate biases (i.e. based on gender, ethnicity, or other sensitive variables), impacting critical areas such as hiring, lending, and healthcare. To address these concerns, recent research explores responsible and trustworthy AI, incorporating non-performance criteria like fairness. This paper employs deonance theory to investigate how the use of conventional and fairness-aware algorithms lead to negative and positive moral reactions among observers. The study employs experimental vignettes involving a fictitious organization's hiring decision, revealing that fairness-aware algorithms elicit stronger other-praising emotions, while conventional algorithms lead to stronger other-condemning emotions. Our research also yields, through empirical evidence, that moral identity influences negative emotional and behavioral reactions, but not positive ones. The findings contribute to understanding how organizational use of algorithms influences external stakeholders' moral reactions, providing insights for managerial decisions. Future studies aim to replicate and extend these findings, exploring a firing situation instead of hiring, identifying thresholds in reactions to fairness vs performance dilemmas, and investigating the influence of personal characteristics on moral responses.