Two Tickets are Better than One: Fair and Accurate Hiring Under Strategic LLM Manipulations

Lee Cohen, Connie Hong, Jack Hsieh, Judy Hanwen Shen

TL;DR: A new model for LLM strategic manipulations grounded in real world hiring systems

Abstract: In an era of increasingly capable foundation models, job seekers are turning to generative AI tools to enhance their application materials. However, unequal access to and knowledge about generative AI tools can harm both employers and candidates by reducing the accuracy of hiring decisions and giving some candidates an unfair advantage. To address these challenges, we introduce a new variant of the strategic classification framework tailored to manipulations performed using large language models, accommodating varying levels of manipulations and stochastic outcomes. We propose a "two-ticket" scheme, where the hiring algorithm applies an additional manipulation to each submitted resume and considers this manipulated version together with the original submitted resume. We establish theoretical guarantees for this scheme, showing improvements for both the fairness and accuracy of hiring decisions when the true positive rate is maximized subject to a no false positives constraint. We further generalize this approach to an $n$-ticket scheme and prove that hiring outcomes converge to a fixed, group-independent decision, eliminating disparities arising from differential LLM access. Finally, we empirically validate our framework and the performance of our two-ticket scheme on real resumes using an open-source resume screening tool.

Lay Summary: Job seekers are increasingly using AI tools like ChatGPT to improve their resumes and cover letters. However, this creates unfairness because not everyone has equal access to these AI tools or knows how to use them effectively. This means some candidates get an unfair advantage, while employers struggle to make accurate hiring decisions because they can't tell which applications were AI-enhanced. We developed a "two-ticket" approach to level the playing field. Instead of just looking at the resume a candidate submits, our proposed algorithm automatically creates a second, AI-enhanced version of every resume. Then it evaluates both versions together - the original and the AI-improved one. This way, all candidates effectively get the same AI assistance, regardless of whether they had access to these tools themselves. We proved mathematically that this approach makes hiring both fairer and more accurate. By giving everyone the same AI boost, we eliminate the advantage that comes from unequal access to technology. We also tested this on real resumes using actual resume screening software and found it works in practice. Our work addresses a growing concern about how AI might increase inequality in job markets while helping employers make better hiring decisions.

Application-Driven Machine Learning: This submission is on Application-Driven Machine Learning.

Link To Code: https://github.com/heyyjudes/llm-hiring-ecosystem/tree/main

Primary Area: Social Aspects->Fairness

Keywords: fairness, strategic classification, hiring, societal impacts, trustworthy machine learning

Submission Number: 7178