In our tests a decade ago, we found that shared autonomy was actually much more frustrating for users than either full or no autonomy. When you were telling the

In our tests a decade ago, we found that shared autonomy was actually much more frustrating for users than either full or no autonomy. When you were telling the computer to go over there and the computer wanted to go over here, it left all of the users frustrated and complaining that the computer wouldn't let them go where they wanted. Maybe you can come up with a better adjudication system than we did, but it is very tricky to design the system so that it catches when the user does something stupid but not when they are doing the right thing- if the computer could figure out the right thing you wouldn't need the human!