Human reliability evaluation for a keyboarding task

摘要

In the majority of human reliability prediction techniques, it is assumed that the probability a human makes an error when performing a task is constant. Similarly, existing system reliability models are based on the assumption that human error rates are constant. This constant rate assumption has received much criticism lately. In this research, a keyboarding task is developed and subjects are recruited to perform the task in an experimental setting. The occurrence of human errors is modeled as a real-time counting process. If human errors that occur at a constant rate, then errors occur according to a Poisson process. Human errors occur during the experimentation are observed and the corresponding data are analyzed for constant error rate behavior using appropriate statistical tools. The data that is not consistent with the constant error rate assumption is modeled as a power law process. Results and conclusions from the keyboarding task are presented.