A Binary LSTM Self-Attention Mechanism Model for Mitigating Academic Ingratiation in the Selection Interview

Abstract

Organisations often face challenges in detecting and preventing ingratiation, a type of intentional selfmisrepresentation during selection interviews that impairs the fairness and accuracy of recruitment decisions. Ingratiation poses substantial risks by allowing the selection of unsuitable personnel, with long-term ramifications for organisational performance and culture. The authors of this paper applied and adapted a deep learning method, specifically a binary LSTM Self-attention model, to mitigate ingratiation during academic selection interviews. As ingratiation manifests itself through strategic linguistic cues and impression-management behaviours comparable to those observed in social media communication, Twitter data serves as an appropriate proxy for modelling such behavioural patterns in an authentic and unconstrained context. Leveraging behavioural data from a Twitter dataset, the model employs deep learning techniques to analyse communication patterns and predict candidate suitability. Combining Long Short-Term Memory (LSTM) networks and self-attention mechanisms enables the model to effectively incorporate complex context-dependent features, thereby enhancing prediction accuracy. This approach not only addresses the limitations of subjective assessment but also aligns with the growing trend of integrating artificial intelligence into human resource procedures. The indicated method was tested on a Twitter dataset, and the findings show that BINSAMLSTM achieved 96% prediction accuracy and 96% F1 score. These results point out to the practical benefits of using this approach, namely: minimising subjective bias and improving consistency in candidate evaluation. While the approach is promising, it also necessitates critical thinking regarding data governance, fairness, and privacy. Compliance with regulatory frameworks such as the Protection of Personal Information Act (POPIA) and the General Data Protection Regulation (GDPR) is critical for maintaining ethical integrity in AI-assisted hiring. This study advances an evidence-based paradigm for mitigating ingratiation, which adds to improving fairness and decision-making in staff selection procedures, particularly in the academic recruiting context