English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Large Language Models (LLMs) have recently emerged as an effective tool toassist individuals in writing various types of content, including professionaldocuments such as recommendation letters. Though bringing convenience, thisapplication also introduces unprecedented fairness concerns. Model-generatedreference letters might be directly used by users in professional scenarios. Ifunderlying biases exist in these model-constructed letters, using them withoutscrutinization could lead to direct societal harms, such as sabotagingapplication success rates for female applicants. In light of this pressingissue, it is imminent and necessary to comprehensively study fairness issuesand associated harms in this real-world use case. In this paper, we criticallyexamine gender biases in LLM-generated reference letters. Drawing inspirationfrom social science findings, we design evaluation methods to manifest biasesthrough 2 dimensions: (1) biases in language style and (2) biases in lexicalcontent. We further investigate the extent of bias propagation by analyzing thehallucination bias of models, a term that we define to be bias exacerbation inmodel-hallucinated contents. Through benchmarking evaluation on 2 popular LLMs-ChatGPT and Alpaca, we reveal significant gender biases in LLM-generatedrecommendation letters. Our findings not only warn against using LLMs for thisapplication without scrutinization, but also illuminate the importance ofthoroughly studying hidden biases and harms in LLM-generated professionaldocuments.

"Kelly is a Warm Person, Joseph is a Role Model": Gender Biases in LLM-Generated Reference Letters