Privacy-Preserving Artificial Intelligence Interview System: A Behavioral Analytics based, Real-Time Integrity Checking, Candidate Holistic Assessment

Conventional techniques of automation of interviews are confined to the analysis of written responses- without paying attention to important behavioral cues and issues of integrity which human interviewers can easily monitor. The current solutions are either based on a cloud infrastructure that poses privacy threats or are not based on the comprehensive assessment capability to offer fair and well rounded evaluation of the candidate. We present a unified model of AI interview that unifies three forms of paradigms for evaluation: (1) context-sensitive questioning, (2) behavioral profiling by face analysis, and (3) automatic integrity testing. The system makes use of privacy protection through locally-deployed language models, behavioural analysis through computer vision. as well as integrity validation via browser-level monitoring. Each of the components is fed into a central scoring Assessment for learning dashboard using visual rubrics. Testing across 150 candidate sessions Improved assessment (88.4% accuracy, 0.87 human agreement) and 94% of integral violations are detected. Behavioral measures provide an additional discriminating power which reduces the number of false positive recommendations by 31%.