Cognition: Memory, Attention, and Learning - Foundations for Designing User-Centered Systems - page 139

interruptions immediately disrupts the main task more than taking the interruption
between subtasks. Dealing with interruptions at scheduled times leads to better
performance on the main task, but poorer performance on the interruptions.
Intermediate strategies offered different trade-offs (McFarlane 1999 ).
Sometimes interruptions are useful. It has been hypothesized that interruptions
are useful for solving hard problems (where there is a so-called incubation effect).
This effect has been noticed by programmers who get stuck on a problem—coming
back to a hard problem later sometimes makes the problem easier. Being inter-
rupted for minutes, hours, or days in these tasks allow the user to forget their
mistakes and mis-starts, or to receive suggestions from the environment. Kaplan
( 1989 ) found that at least part of the incubation effect came from cues in the
environment. One problem he gave subjects was ''What goes up a chimney down
but not down a chimney up?'' Subjects who were called and asked by Kaplan if he
had left an umbrella in their office were much more likely to solve the problem, but
did not attribute the cause to his call.
As you consider the larger context of your users' tasks, you should keep in mind
the possible effects of interruptions. Some interfaces will be less sensitive to the
effects of interruptions. We are just starting to be able to predict and measure these
effects.
5.3.6 Automation Deficit: Keeping the Human in the Loop
If you want to get funny looks, while you are driving a car, ask your passenger
''Do you ever wake up and find out you are driving and wonder where you are
going and where the other cars are?'' Some find this humorous, and others, if they
do not know you well, may be startled. This is similar to what happens when
someone has to resume or take over a task at short notice when the system
relinquishes control because it does not know what to do. This happens to people
such as aircraft pilots and power plant operators who find themselves out of the
loop, i.e., not being kept up to date by the systems about what it is doing. The user
will then allocate their attention to dealing with the task, but they have to spend
time and effort trying to understand the current situation before they can diagnose
the problem and resolve it.
Generally, the way to avoid this problem is to include more status information
to keep the user continually informed before the situation requires their input. This
allows one to develop and maintain a mental model of that situation and how it is
unfolding, so that users can decide when they may need to take action.
This whole argument about the need to keep the human involved in the control
loop was first summarized by Bainbridge ( 1983 ), and it remains as an issue today
(Baxter et al. 2012 ). Woods and Patterson ( 2001 ) explain some of the effects on
users if the technology does not keep them kept informed about what it is doing.