Operator situation awareness monitoring in the cognisant control room

The use of a suitably expressive underlying situation model allows the detection of differences in the situation awareness of an operator as compared to a situation model estimated from low-level tracks. A grammatically based situation model allows efficient construction of these models. Operator speech and actions can be interpreted as statements about the situation and expressed using the same model. Inconsistencies between the operatorpsilas and the systempsilas view can be flagged to the operator as alerts. Using a sufficiently high-level situation model means alerts can signal errors in force structure and tasking that could not be generated from raw track data. A multi-mode interface limits interruptions by alerts by gathering as many clues as possible to the operatorpsilas awareness from his speech and GUI interactions. For a simple scenario, false force structure alerts due to situation estimation errors are shown to occur, on average, less than once per hour.