Mitigating Human Bias in Candidate Evaluation Through an AI-Driven Multimodal Assessment System

Abstract

The recruitment process is often marred by human bias and inconsistent evaluations, especially when relying on traditional interview techniques. This paper introduces a scalable AI-driven, multimodal interview system designed to deliver objective and comprehensive assessments of job candidates. The proposed framework integrates natural language processing, computer vision, and code quality analysis to evaluate both technical skills and interpersonal attributes. It features four key components: (1) automated skill and professionalism assessment through resume parsing and behavioral analysis during a mock exam, (2) voice-based confidence evaluation using speech feature ex-traction and integrates natural language processing, (3) a gamified technical interview environment with real time stress detection via facial expression analysis, and (4) code complexity and maintainability analysis employing Cyclomatic Complexity, Cognitive Function Complexity, and Weighted Code Complexity metrics. Experimental results demonstrate that the system provides accurate, bias-reduced evaluations across technical and non-technical domains. By unifying behavioral and technical metrics, this approach offers a fair, efficient, and data driven alternative to conventional hiring methods.