Institutional Prestige Bias in Large Language Models
August 20, 2026
Experiments across four LLMs show a statistically robust institution-tier gradient of +0.297 in candidate evaluations. The research finds that journal prestige exerts a significantly stronger influence on model bias than the candidate's country of origin.
HOW THIS AFFECTS YOU
●
policyYou must account for systematic prestige bias when deploying LLMs for recruitment or evaluation.