On the Problem of Human Bias in the Management of the Organization's Personnel

Abstract

In recent years, of all the resources involved in the organization in the implementation of production processes or the provision of services, managers have paid more attention to human resources. Experts argue that investing in human capital generates more revenue for companies than even modern technology. Therefore, to achieve optimal efficiency and productivity in today's highly competitive environment, organizations must use effective methods of managing human resources. In the organization's personnel management system, when recruiting personnel, a subjective approach takes place when evaluating the applicant accepted for a vacant workplace in the organization. The aim of this study is to use modern innovative deep learning models to eliminate recruitment manager bias by increasing the flexibility of filtering parameters by implementing an adversarial autocoder. A person's bias in the personnel management process, in particular when hiring managers inadvertently favor candidates who have similar experiences or values to their own, is a recurring issue that businesses often deal with. The hypothesis of the study is that, having studied the personnel management model in the Kazakhstan branch of the audit and consulting company KPMG in Almaty with 25 years of experience with more than 100 employees, it will be possible to more accurately assess the applicant's participant through an adaptable scale factor of various parameters, including career progress, organizational features and potential risks. Deep and adversarial autocoders were evaluated and contrasted with the Naive Bayes reference classifier. The data were pretreated (cleaning, merging, denoising) as well as optimized for the neural network used in deep learning. Metrics derived from the confounding matrix were matched and the results showed that the adversarial autocoder produced better results and that deep learning models tended to be superior in terms of digital prognostic analysis.