A Fair Candidate Screening System’s documentation

Introduction

In this work, we propose a fairness-aware AI-based system for candidate screening, grounded in the analysis of a real-world dataset provided by the Adecco Group. The dataset consists of AI-generated candidate-job matches enriched with demographic and educational information. Our approach aims to detect and mitigate structural and cognitive biases that may arise during algorithmic matching processes. We begin by analyzing the data distribution to identify patterns of disparity, and subsequently evaluate the fairness of the existing recommendation mechanism using established bias metrics. To promote more equitable outcomes, we explore preprocessing strategies and fairness-aware ranking techniques.

Contents:

• Dataset and use-case context
• Methodology
  • 1. Data Cleaning (Selection)
  • 2. Data Loading & Understanding
  • Sensitive feature selection
• Fair Candidate Screening System
  • Setup
  • Dataset Loading
  • Data Preparation
  • Data Preprocessing
  • Bias detection
  • Bias mitigation techniques