Algorithmic Path Dependence in Linked HR Decision Systems

This study examines how interconnected artificial intelligence (AI) tools can turn early human resource (HR) allocations into evidence for later career decisions. A counterfactual-twin Monte Carlo simulation compares isolated and linked systems, alongside periodic record resets and opportunity equalization. Paired employees share fixed capability and random shocks but receive slightly different initial scores. Linking decisions amplifies this initial difference through opportunity allocation, observed performance and reusable records. Periodic resets reduce final-career sensitivity, while opportunity equalization affects access more strongly than terminal outcomes. Near parity in average group outcomes can coexist with positive sensitivity among matched individuals. Additional tests preserve the direction of the architecture and intervention contrasts across the examined coefficient settings, although their magnitudes vary. The findings motivate evaluating decision lineage, opportunity access and record reuse across HR systems. They identify mechanisms within a synthetic, uncalibrated model and do not estimate effects in real organizations.