A HUMAN DECISION PREDICTION MODEL FOR RECOGNIZING LOSS OF CONTROL IN PROCESSES

Abstract

This article presents the results of a study on human decision-making patterns with the aim to analyze their impact on the effectiveness of automated control systems. The findings of this study could be applied in the development of decision support systems, as most of the cases are influenced by such factors as complexity of the situation, operator experience and time limitations. This research was conducted with the use of chess data, as this domain provides well-defined metrics for the required parameters.The study evaluates the importance of each group of indicators that have an impact on human decisions. The analysis was based on chess games in which one of the players could resign. The resignation decision in chess could be perceived as loss of control. In other domains it could be seen through problem avoidance, decision delay, or responsibility delegation.Methods of correlation analysis were used for evaluation of the connections between variables. Machine learning methods such as neural network, random forest and gradient boosting XGBoost were used for prediction of human resignation decisions. The key factors of the decision-making process were determined by analysis of impact of input parameters using feature importance techniques such as gain and permutation importance.The outcome of the research was the creation and training of models to predict human decision about loss of control or hopelessness. The performed analysis demonstrated that complexity of the situation is by far the most important factor.Other indicators such as person’s experience, time limitations or motivational level have some impact, but less significant.