Abstract
Many organizations actively employ artificial intelligence (AI) tools to enhance the speed, scale and consistency of decision‐making, including the resource intensive procurement process of supplier scouting—identifying and selecting high‐quality suppliers. This process involves high‐stakes procurement decisions that shape the structure and performance of supply networks. However, over time, outcomes from AI‐aided supplier scouting often show a gradual drift as supplier diversity narrows, novel entrants are overlooked and recommendations increasingly mirror previous selections. This paper unpacks this problematic drift by constructing a middle‐range theory of bias co‐evolution grounded in ecological rationality, which defines decision quality as the fit between decision strategies and decision environments. We posit two second‐order biases— status‐quo bias in human judgement and machine‐feedback bias in algorithmic outputs—that mutually evolve across repeated decision cycles, progressively reinforcing one another. This co‐evolution creates a temporal trajectory where bias becomes increasingly embedded in both human and algorithmic decision systems. We formalize this theory through nine propositions linking market, data, algorithmic and human decision structures. In closing, we offer four design principles—monitor bias transitions, disrupt feedback loops, isolate decision parameters, create algorithmic circuit‐breakers—to preserve decision fit and enable responsible AI‐aided decision‐making.

This publication has 108 references indexed in Scilit: