Developing A Fair Hiring Algorithm Using LLMs

Developing A Fair Hiring Algorithm Using LLMs

Ashutosh Marathe, Atharva Thakur, Shruti Dhumal, Sidhant Gokulpure, Aryan Shisode, Shreya Basarikatti

Department of Multidisciplinary Engineering (AI & DS)

 Vishwakarma Institute of Technology, Pune, 411037, Maharashtra, India

Abstract - In this project, our primary objective is to revolutionize hiring practices by harnessing the potential of advanced technologies, specifically Large Language Models (LLMs). We strive to develop an impartial and comprehensive hiring algorithm that addresses biases and advocates for diversity, ultimately fostering a more equitable workforce and enhancing overall organizational outcomes. The methodology encompasses two crucial phases: data collection involving diverse candidate resumes and unbiased job descriptions, followed by leveraging LLMs and LangChain for pre-training and fine-tuning using carefully curated unbiased training data. Custom prompts are carefully designed to guide the algorithm's decision-making process during candidate evaluation. The algorithm's fairness was evaluated through bias detection, equal opportunity assessment, and adverse impact analysis, ensuring equitable outcomes, establishing the foundation for a fair hiring algorithm and promoting inclusivity in recruitment.

Keywords - LLM, LangChain, Python, Agent, Chains.