Search papers, labs, and topics across Lattice.
This study investigates the impact of name-conditioned evaluative framing in resume summaries generated by LLMs, focusing on how such biases manifest in hiring contexts. Analyzing nearly one million summaries from four models, the authors find that while the factual content of resumes remains stable, the evaluative language varies significantly based on the race-gender identity implied by names, particularly in open-source models. The research reveals that this evaluative instability can lead to asymmetric biases in hiring simulations, suggesting that conventional fairness audits may overlook critical issues in LLM outputs.
Evaluative language in LLM-generated resume summaries can introduce significant biases that traditional fairness audits might miss, destabilizing hiring processes.
Research has documented LLMs'name-based bias in hiring and salary recommendations. In this paper, we instead consider a setting where LLMs generate candidate summaries for downstream assessment. In a large-scale controlled study, we analyze nearly one million resume summaries produced by 4 models under systematic race-gender name perturbations, using synthetic resumes and real-world job postings. By decomposing each summary into resume-grounded factual content and evaluative framing, we find that factual content remains largely stable, while evaluative language exhibits subtle name-conditioned variation concentrated in the extremes of the distribution, especially in open-source models. Our hiring simulation demonstrates how evaluative summary transforms directional harm into symmetric instability that might evade conventional fairness audit, highlighting a potential pathway for LLM-to-LLM automation bias.
