Predicting honest behavior based on Eysenck personality traits and gender: an explainable machine learning study...

Area of Science:

• Psychology
• Aviation Safety
• Machine Learning

Background:

• Limited research applies Eysenck personality traits to honesty prediction in aviation.
• Dishonest acts in aviation, like incident underreporting, have severe safety implications.

Purpose of the Study:

• To investigate Eysenck personality traits (Neuroticism, Psychoticism, Extraversion) and gender as predictors of dishonest behavior in aviation.
• To apply explainable machine learning to model these relationships and identify key honesty predictors.

Main Methods:

• Behavioral data collected from 102 flight and air traffic control cadets using a coin-toss task.
• XGBoost (explainable machine learning) modeled nonlinear relationships between personality, gender, and honesty.
• SHAP analysis identified key predictors and model performance was evaluated via AUC.

Main Results:

• XGBoost demonstrated high predictive accuracy (AUC = 0.802).
• Neuroticism emerged as the strongest predictor of dishonesty.
• Males exhibited higher dishonesty rates, and threshold effects were observed for Psychoticism and Extraversion.

Conclusions:

• This study systematically applies the Eysenck model to aviation honesty prediction.
• Gender significantly moderates honesty tendencies.
• A SHAP-driven framework links machine learning insights to psychological theory for data-driven cadet screening and aviation safety interventions.