Adaptive automation is an area where we’re looking at things that when we’ve got humans and machines interacting together, increasingly you’ve got things that are automated, whether they be computers, UAVs, things like that. So adaptive automation is looking at smart ways to make the automation adapt to the environment so that it can help humans interact better with the machines that they’re interacting with. One of the things that we saw that was lacking in adaptive automation was the fact that a lot of automation research focuses on creating automations that perform their tasks as best as possible. So for example, if we’re trying to fly a UAV, we want to perform the portion that automation is supposed to take over…we want to perform that as best as we possibly can. But the problem with that is when you start involving humans and machines together, when the machine operates as best it can, sometimes that hurts the overall human-machine team because it will confuse the human or take over too much of the responsibility for the human. So what we tried to do is to take that a step further and try and create automations that improve the overall human-machine team by improving the way the human can interact with the machine. The way we tested this was we were able to take a tablet computer game and let users interact with a system that they work with every day. We were able to take a system where people would draw trajectories for space ships that are coming onto the screen and then we would automate a portion of that task. By seeing how people interacted with those automations, we were able to see how the slight changes that we make when we’re automating a task, how those slight changes can cascade when the human—and different types of humans—are interacting with the same systems. A few of the interesting findings that we saw is pertaining to trust. When people are working with computers, they really care about how much they trust a computer. I can create the best automation in the world, but if the person can’t both understand how it works and the decisions it’s making and predict what it’s going to do, they’re not going to trust it and they’re not going to use it well. One of the really interesting things that we can push for the Air Force that some of this research gives us the ability to do is it helps us to design better automation for people to use. And not just in things like UAVs, but also just in like day-to-day tasks. Once of the things that AFIT and the ANT Center specifically has the capabilities to do that some of the outside places can’t do is, we have unique resources where we can take problems that we can test out in toy environments and we can then apply them in real-world data sets. We can actually go to people who are using human-machine teams who are interacting with computers and we can test out some of our ideas and we can push those to real-world operators.
