Adaptive selection of personality items to inform a neural network predicting job performance

Abstract

Connectionist or "neural" networks, developed as a model of cognition, are also a general statistical model with practical applications. Adaptive testing, traditionally based on item response theory, is a way to improve the efficiency of a test. A hybrid system is developed that captures the main advantages of both technologies: the modeling flexibility of a neural network, and the efficiency gains of adaptive testing. A prototype is implemented for the case of a personality assessment used to predict job tenure at a national retail chain. Applicants' assessment and subsequent employment data are used to demonstrate the prototype's effectiveness.