Experiments on the Use of Bayesian Inference and Evaluation of Human-Machine Interfaces on Situation Awareness

摘要

The Bayesian inference has been widely used in various research fields. It is a method of inference that uses Bayes' rule to newly estimate the probabilities of certain hypotheses when additional evidence is acquired. According to certain recent studies, malfunctions of nuclear power plants can be detected using Bayesian methodology by consistently piling up new incoming data and updating estimations. However, these studies tend to assume that people behave just like computers. Such an unfounded assumption may create a problem in real world applications. Studies in cognitive psychology have indicated that, as the amount of information increases, it becomes less likely for an individual to retain the information in its entirety. This is due to two shortcomings of the human brain: the limited capacity of the working memory and the inevitability of attention lapses, which often interfere with our information processing. Due to these human factors, ideal and real operators show different results in terms of situation awareness. The purpose of our research is to consider how actual human behavior contrasts with our expectations, and how such disparity, if at all, affects the results of the conventional Bayesian inference. We will compare an ideal operator with a real operator and, for the real operator, we will use both text-based HMI and infographic-based HMI to further compare two existing real operators.