Task level errors for human error prediction in GOMS-HRA

Abstract

Goals-Operators-Methods-Selection rules (GOMS) was originally developed as a task analytic tool for modeling behavioral primitives in human users of human-computer interfaces. GOMS was recently adapted for Human Reliability Analysis (HRA), producing the GOMS-HRA method. The GOMS-HRA method provides a taxonomy of task level primitives in human activities that correspond to human error probabilities and task timing. The GOMS-HRA method has been used in computation-based HRA (CoBHRA), due to its calibration to the subtask level of human performance, the optimal decomposition level for dynamic risk modeling. GOMS-HRA has also been linked to procedures, and it is possible to map procedure steps (called procedure level primitives) to task level primitives. This paper introduces another important development to the GOMS-HRA framework—the task level errors, which represent the use of the GOMS-HRA taxonomy for predicting human error types. While many HRA methods map task types or task primitives to error rates, the prediction of error is often a generic error type. In reality, each task level primitive has predilections to certain types of errors. To model human error dynamically requires the determination of the types of errors that can occur.