727 simulator around the streets of San Francisco, a commercial airline pilot at
the controls, flying around the Transamerica building. Oops, we flew through
the Transamerica building. Not even a tremble. Once we dove into the ground
a close to supersonic speeds. Those of us in the cockpit felt somewhat nau-
seous: Our minds expected sights, sounds, and movements that did not occur.
The computer simulator just kept going. Buildings, walls, even the ground are
just numerical and graphic abstractions: To a simulator, there is nothing im-
possible about being 1 meter below the ground.
Suppose a programmer of computer games developed an exploratory game
with Harjime ́, the protagonist, wandering through the halls of the enchanted
castle. Writing the part of the program that controls Harjime ́isn’t all that diffi-
cult, nor is the part of the program that simulates the castle. Want to simulate
the room with the hidden treasure? Just draw in the locations of the walls, fur-
niture, secret keys and panels, and the hidden door. But making the simulation
workwouldbeacomplextask.ThehardpartistomakesurethatHarjime ́
doesn’t walk through the physical objects in the room. If Harjime ́picks some-
thing up and then puts it down somewhere else, the programmer has to worry
about whether there is a supporting structure at the new location, or if the ob-
ject will fall, tilt, or slide. Harjime ́’s movements would also have to be carefully
monitored to make sure there was always a supporting floor or surface. Har-
jime ́couldn’t move up or down unless there was always a suitable support
(but he would have to move up, down, or around when he encountered stairs,
ramps, furniture, and elevators). Although the task of programming Harjime ́
andthecastlecouldbegiventonovices,thetaskofprogrammingtheinterac-
tion of the two is complex and difficult enough that it would tax even the ex-
perts: How quickly the program could actually execute would be determined
by how well it could compute the necessary constraints and interactions.
The point is that in the real world, the natural laws of physics allow only the
appropriate things to happen. There is no need to compute whether you are
walking through a wall: You simply can’t do it. In the artificial world of com-
puter simulation, much of the computational effort goes into the part that
results from the artificiality of the situation.
It has long been noted that in dreams, people are free of the constraints of
everyday life. We can visualize doing things that are impossible in the real
world. Ah, the freedom of dreams, the fantasies released. The impossible ac-
tions of dreams might be ways by which people satisfy their fantasies. But they
might also result from the impoverished programming power of the human
mind.
Suppose, just suppose, that the wonderfully creative fantasies of our dreams
are artifacts, accidents of the fact that our minds can’t quite handle the com-
putational job of doing accurate simulation. Adream, after all, is a simulation
of human action within a simulated environment. The simulation program
is executed within the human mind—a disembodied mind, however, for the
sense organs are inhibited and the voluntary muscle system inoperative. Con-
siderwhatitwouldtaketorunthissimulationproperly.Thepeopleand
objects would have to be created and their actions determined. The environ-
ment would have to be created. And finally, the interactions among all the
450 Donald A. Norman