Predicting Honest Behavior Based on Eysenck Personality Traits and Gender: An Explainable Machine Learning Study Using SHAP Analysis

Yu Meng ^* Zili Peng Zhitong Zhang Qiaolin Chen Hanxi Huang Yihan Chen Mengqian Zhao

• Civil Aviation Flight University of China, Guanghan, Sichuan Province, China

This study addresses a critical gap in aviation safety research by investigating the predictive role of Eysenck personality traits (Neuroticism, Psychoticism, Extraversion) and gender in dishonest behavior, particularly within high-risk aviation contexts. While prior literature has explored personality models like the Big Five and HEXACO in ethical decision-making, empirical studies utilizing the Eysenck framework to predict honesty remain scarce, especially in aviation-a field where dishonest acts (e.g., underreporting safety incidents) pose severe public safety risks. With a coin-toss task, we collected behavioral data from 102 flight and air traffic control cadets and employed explainable machine learning models to uncover nonlinear relationships between personality, gender, and honesty. Results demonstrated that XGBoost outperformed other algorithms (AUC = 0.802), with SHAP analysis revealing neuroticism as the strongest predictor of dishonesty, alongside significant gender differences (males exhibited higher dishonesty rates). This research contributes three key advancements: (1) it extends the application of the Eysenck model in aviation honesty prediction, addressing its underutilization in existing literature; (2) it integrates gender as a critical moderating variable, enhancing predictive precision; and (3) it establishes a transparent, SHAP-driven framework that bridges machine learning's "black-box" limitations with psychological theory. Practically, these findings provide aviation authorities with a datadriven tool for screening cadets' honesty tendencies during recruitment and training, ultimately supporting safer operational environments through targeted psychological interventions.