LLMs Exhibit Covert Dialect Bias in Housing-Related Judgments
September 17, 2026
Analysis of ten open-weight LLMs reveals that while explicit racial bias is often mitigated, covert linguistic profiling persists in internal probability distributions. The study shows biased social judgments across Standard American English, AAVE, Nigerian Standard English, and Nigerian Pidgin in housing contexts.
HOW THIS AFFECTS YOU
●
researcherThis highlights a critical gap in current alignment techniques regarding covert attitudinal associations.
●
policyYou should be aware of hidden biases when deploying LLMs for high-stakes screening tasks.