Developing a Bias-Resistant AI-Powered Conversational Interview System for University Admissions: Integrating Fairness-by-Design with Real-World Validation
Problem Being Solved
The pursuit of equity and efficiency in university admission processes forms the backbone of academic integrity and societal progress. However, these processes face persistent challenges: traditional interviews demonstrate substantial variability in evaluator ratings even for identical candidates, with factors like interviewer gender and panel composition influencing scores significantly. As institutions struggle with increasing application volumes and complexities of holistic review, the potential of Artificial Intelligence to streamline operations becomes increasingly attractive yet AI-based systems intended to support decision-making have introduced their own challenges, from embedded demographic bias to perceptions of unfairness and lack of transparency. My commitment to this research stems from discovering embedded regional bias in Nigeria's Poverty Alleviation beneficiaries' selection program. The AI system systematically discriminated against qualified individuals revealing how well-intentioned AI can amplify existing inequalities without proper fairness frameworks. Subsequently, navigating my own PhD admissions process exposed systemic issues including unexplained rejections and potential implicit biases that overshadowed objective qualifications. This dual experience with algorithmic and human bias drives my commitment to designing transparent, auditable, and accountable AI systems that eliminate subjective discrimination while providing timely and fair evaluations. Primary Research Question: How can a bias-resistant AI-powered conversational interview system be designed to enhance fairness, efficiency, and accuracy in university admissions? Supporting research questions examine: (1) What scalable conversational AI architecture with tiered processing and multilingual capabilities enables natural, adaptive interviews while maintaining real-time fairness monitoring? (2) How can comprehensive bias detection and mitigation frameworks be embedded directly into conversational systems to ensure fair evaluation across demographic dimensions? (3) How can AI-generated assessments be validated through longitudinal tracking, peer assessment, and multi-stakeholder feedback while ensuring privacy protection? (4) What implementation strategies enable successful deployment in existing university admissions workflows? This research addresses three critical gaps: the lack of comprehensive conversational AI with integrated fairness mechanisms for admissions, the absence of real-time bias detection in interview systems, and limited validation of AI conversational assessments against educational success predictors. By developing actual AI systems with embedded fairness monitoring and deploying them for real-world validation, this research produces both technical innovations and practical frameworks for responsible AI deployment in high-stakes educational evaluation.
Keywords
AI NLP BIAS FAIRNESS EDTECHProof of Registration
SHA256 hash generated at registration time. Use this as timestamped proof of your research topic.
41a679458d5cd97dbd85fde0da331b791e3f1056020d0b10585fe74beae548e4