In a previous post, I described 4 ways Discovery Machine models human behavior. The third way is by building models that multi-task to efficiently complete goals.
Because we wanted our behavior models to represent human performance, we began by researching how people handle completing competing goals. A chapter on attention and multi-tasking in Modeling Human and Organizational Behavior: Application to Military Simulations describes 4 ways in which people perform with a high cognitive load (i.e. multiple competing goals):
1. They continue doing everything but less well
2. They put tasks in a priority queue
3. They reduce the number of things being attended to
4. They drop everything and “walk away”
We considered using each of these methods in our crew member behavior models and ultimately decided that models incorporating the second option would provide the best training.
As I’ve described in previous posts, our behavior models reflect on process models to formulate the set of goals to follow to complete their mission; they may also receive incoming messages, leading to the formulation of new goals. Once these goals have been created, our behavior models prioritize them based on their completion value and estimated time to completion.
In a multi-tasking study cited by Kushleyeva et. al., “subjects were more likely to switch away from a task whenever continuing with that task required them to perform a discrete, time-consuming subtask.” We used this research to determine when our behavior models would switch tasks.
This can be seen in the following branch of one of our crew member models:
This branch is waiting for one of its assets to drop a buoy before disseminating that information to another crew member. Before it begins the act of waiting — what Kushleyeva et. al. would describe as a discrete, time-consuming task — it refers to its goal set to see if another goal can be completed. If such a goal exists, it completes it and then resumes waiting for the buoy to drop.
The same behavior model looks like this at a higher level:
The model concurrently performs the goals it has set for itself while monitoring the situation. The situation awareness branch is based on Endsley’s research on the three levels of situation awareness. The creation and prioritization of future intentions (i.e. goals), two of the three major skills Burgess et. al. have found to be crucial for successful multi-tasking performance, occurs in this branch. The third skill Burgess et. al. have identified to be essential, the ability to switch from carrying out one intention to another when appropriate, occurs in the ‘Perform Goals’ branch.
In my next post, I will describe how our crew member behavior models project potential courses of action to make nuanced decisions.
If you would like to learn more about Discovery Machine’s human behavior modeling approaches in action, download our free whitepaper about using adaptive entity behaviors in training.