[Bulk] Re: Imagination and
control
[Martin Taylor 2006.03.12.16.42]
[From Rick Marken
(2006.03.12.1015)]Martin Taylor (2006.03.10.10.34)
But, and here’s the problem,
“Imagination Mode” is ordinarily treated as a feedback path
in a perceptual control loop. It is generated by a simulation of what
the consequence would be if one performed a certain act.Actually, an imagined perception is
modeled as the playing of the reference signal right back into the
perceptual path of the system sending that reference signal. So an
imagined perception is just perceiving what you want.
That simply can’t be right. I can and often do imagine all sorts
of horrible possibilities that I don’t want, and if I have any control
over them, I act so that the things I imagine don’t happen. Whatever
else they may be, they certainly aren’t played back reference
signals!
Having just written what I said above, it seems clear
to me that the World Model in the feedback loop of a control system is
not the only World Model – or else it’s in the wrong place. The World
Model that predicts the effects of one’s control actions to produce an
imagined result must also include the predictable changes inthe world
that are not affected by one’s own actins. One’s actions are imposed
on what the world would do if we abstained from control, which is the
situation discussed above.I’m going to be doing a section on
imagination soon in my Cognition course so this little discussion has
been helpful. One of the most influential studies of imagination is
the Shepard/Metzler “Mental rotation” study. What you seem
to have her is a situation where, in order to determine if two
pictured objects are the same or not, you imagine what would
happen to one of the objects if it were actually rotated. You
don’t actually rotate the object, so the feedback path connecting your
actions (applying a rotation) to the resulting perception (or a
rotated form) does not go through the environment. So you are
controlling a lower level perception (of rotation) in imagination as
the means of controlling for the higher level perception of a
relationship (sameness) between the pictured objects. I think it’s
interesting that people seem to do, in imagination, what they would do
with the objects themselves (rotation) in order to determine if they
are the same or not. This is interesting because you might imagine
that people could instantly imagine the object in the rotated state in
order to make the comparison. After all, there are no physical
constraints on the object in imagination. But Shepard and
Metzler show that the time required to evaluate the relationship
between objects is proportional to the angular difference between the
objects, which suggests that the imagined rotation is constrained in
the same way as the physical rotation; it takes longer to mentally
rotate in object through 160 degrees than it does to mentally rotate
the object through 60 degrees, justas it necessarily takes longer to
physically rotate an object through 160 as compared to 60
degrees.The textbook I am using describes other
studies that also suggest that imagined means of achieving higer-level
goals are constrained in the same way they would be if the means
(lower level perceptions) were controlled via the feedback path. For
example, answering questions about higher level properties of objects
(like “does the tiger have spots?”) takes longer if these
objects are imagine as being confined in small rather than as large
spaces. As in the “perception” mode it is harder to
“see” the properties of small as compared to large
objects.These results suggest to me that the
hierarchical control of imagination is similar to the hierarchical
control of perception, except that, when controlling in imagination,
there are no disturbances that occur when you are controlling the
lower level perception as the means to controlling the higher level
one.It seems to me that there should be some
variation of the Shepard/Metzler study that could be done based on an
understanding of their results in terms of control theory. I’d be
interested in hearing what others, particularly the experimentalists,
think about this.
I quoted the above in its entirety, because almost all of it
seems right. My 'almost" is because I’m not at all sure about the
“no disturbances” aspect.
I think that much of your discussion (with which I otherwise
agree) argues against the notion that imagination is a direct playback
of a reference signal at any level of the hierarchy. It is, however,
consistent with the playback of an output signal through a World
Model.
Martin