Description
Represents a system's missions as a network of tasks with sampled times and error probabilities, then runs it many times to predict crew and system performance under alternative designs.
Task network models are built from a task analysis: each task node has a time and error distribution (for example normal or exponential, with mean and SD), a release condition that holds it until met, and effects at its beginning, launch and end that change the system state. Branching after a task can be probabilistic (a weighted random draw), tactical (the highest-value next task) or multiple (several tasks start at once). Unlike a single-pass human reliability model such as THERP, the network is run many times to give a distribution of mission time, success and errors. Meister notes such models are human constructions but valuable for studying molar variables that cannot be manipulated in real systems, such as crew size, task allocation, display formats or workload, for instance whether a destroyer could be run by 95 sailors instead of 300. Their credibility depends on validation against real performance and on the quality of time and error inputs.
Attributes
- Discipline
- Task Analysis & Cognitive Modelling
- Loss type
- Siloed
- Data
- Analytical / expert
- Effort
- Months
Steps
- Study the task analysis of the real system and decompose it into mission, functions, tasks and subtasks to the depth needed.
- Define the network: the relations among tasks, release conditions and branching rules (probabilistic, tactical or multiple).
- Select a simulation engine (for example Micro SAINT or IMPRINT, or a general discrete-event tool).
- Specify inputs: task time and error distributions, interaction rules, equipment data and operator assignments.
- Run the model many times (hundreds to thousands of replications), compare outputs with performance data from the real or a similar system, and refine until they match reasonably.
- Use the verified model to compare design alternatives such as crew size, automation, task allocation or procedure changes.
Sources
- Conceptual Foundations of Human Factors Measurement (Meister, 2004)
Also known as
- SAINT / Micro SAINT task network simulation
- Reductionistic human performance system model
- Task network model
- IMPRINT
Categories
- User Research
- Task Analysis & Cognitive Modelling