TY  - JOUR
T1  - Artificial Intelligence and the Reproduction of Organizational Inequality
A1  - Charlotte Dupuis
A1  - Hugo Perrin
A1  - Elise Morel
A1  - Thomas Martin
JF  - Journal of Applied Organizational Systems and Behavior
JO  - J Appl Organ Syst Behav
SN  - 3108-852X
Y1  - 2024
VL  - 4
IS  - 2
DO  - 10.51847/3o8nZO48aI
SP  - 93
EP  - 102
N2  - Artificial intelligence is increasingly used to allocate work, evaluate performance, screen applicants, distribute opportunities, and structure managerial attention. Yet organizational debate often isolates technical bias from implementation, worker participation, error distribution, and accountability, obscuring how disadvantage can persist across an entire decision process. This narrative review examines how artificial intelligence may reproduce organizational inequality through four connected domains: bias in data and system design, participation deficits during implementation, unequal exposure to automated error, and accountability failure. A structured, evidence-bounded review approach combined explicit questions, relevance-focused searches, transparent screening, critical reading, thematic comparison, and PRISMA-compatible reporting. The literature converges on the importance of data provenance, proxy and objective choice, authority allocation, information asymmetry, contestability, and remedy capacity. It conflicts, however, over whether observed disparities arise primarily from encoded social history, optimization processes, institutional conditions, user interpretation, or pre-existing organizational inequality. Evidence is also heterogeneous across computational audits, field studies, experiments, qualitative process research, conceptual analyses, and review articles. The article therefore proposes an original multistage synthesis in which design choices shape implementation participation; participation shapes exposure and challenge capacity; accountability conditions correction; and organizational responses may feed into later data and evaluation. This synthesis is not a validated causal model. Its implications are conditional on occupation, employment relation, institutional context, worker resources, organizational governance capacity, and time. Longitudinal, multilevel, comparative, and intervention-based research is required to test, refine, or reject the proposed relationships.
UR  - https://apsshs.com/article/artificial-intelligence-and-the-reproduction-of-organizational-inequality-a45zd2r6ukbnh7c
ER  -