Candidate assessment in asynchronous video interviewing: a hybrid hyperbolic fuzzy integrated MCDM framework with generative AI expertise

Abstract

Asynchronous Video Interviewing (AVI) is increasingly used in large-scale recruitment; however, its evaluation remains constrained by opaque predictive models, limited interpretability, and reliance on subjective human judgement. Interview data is often vague and complex. In this study we propose a novel “Multi-Criteria Decision Making (MCDM)” framework designed to navigate this uncertainty by identifying the underlying influence relationships that drive AVI assessment outcomes. The framework integrates “Hyperbolic Fuzzy Sets (HYFS)” with a multi-perspective “Decision Making Trial and Evaluation Laboratory (DEMATEL)” approach to estimate criteria weights and identify directional influence structures while employing state-of-the-art generative AI models as scalable AI-based evaluative agents. To address variability and potential bias in human ratings, a multi-attitudinal variance-based method is used to compute panel participant weights, and interviewees are subsequently ranked using a HYFS-integrated “Weighted Aggregated Sum Product Assessment (WASPAS)” technique. The framework is demonstrated on a real-world AVI dataset comprising 138 interview recordings evaluated across 18 behavioural and non-verbal criteria. The results show that Eye Contact, Smiled, Speaking Rate, Friendly, Paused, and Calm act as dominant drivers, while micro-level speech disfluencies such as fillers exhibit comparatively lower influence once broader behavioural cues are considered. Overall, the proposed framework yields rankings aligned with benchmark scores while offering enhanced interpretability and directional influence insight, supporting fair and trustworthy automated hiring systems.