Control of Imagination (was Facts and Theories)

[rsmarken]

[From Rick Marken (2014.11.29.1030)]

···

Martin Taylor (2014.11.28.11.48)–

MT: Very handsome of you, and I thank you for it, while adding my own

apology for participating in the conflict escalation.

MT: With luck, we should be able to terminate this and the ancestor

thread at this point.

RM: OK, then why don’t we start a new thread based on your proposal about controlling perceptions partially in imagination. In PCT as it now stands control is either of sensory based perceptions – perceptions of aspects of the environment – or control of imagined perceptions – perceptions that are replays of reference signals retrieved from memory (as per Figure 15.3, p. 223 of B:CP, 2nd Ed). So in the original version of PCT, whether we are controlling perceptions or imagined perceptions was conceived of as an all or none process – the result of throwing the neural “switches” from one connection to another.

RM: You have proposed (and I know that we have discussed this on CSGNet before so I know that Bill agreed) that it is possible that the perceptual signals that are controlled could be partially the result of sensory based and partially the result of memory based inputs to a perceptual function. The memory inputs would be there to “fill in” what might be missing from the sensory based input (due to brief obstructions or whatnot).

RM: I think it would be a great contribution to PCT if you could determine how (and, of course, whether) this mixed sensory/memory perception concept should be incorporated into the PCT model. I just ran a quick little simulation where I had systems controlling a sensory based perception p = aX+bY and switched X from being an environmental variable to being a reference signal from memory(an imagined input, call it X’) and what I found was that control of the now partially imagined perception – p = aX’+bY – remained intact but the environmental correlate of that perception – aX+bY seemed to be completely uncontrolled. I haven’t done the quantitative calculations yet to determined whether control of aX+bY is really lost completely when only X is imagined, so it may be true that control of aX+bY is not completely lost when one component of the perception of that variable is imagined. But this would be a nice thing to determine to see how control of a partially imagined perception might work.

RM: Another aspect of this that might be good to look into is whether having an imagined component of a perception improves control when the sensory component is lost, at least for a short time. I think you have suggested that this may be the case and I think it’s is possibly true. For example, suppose you have a person do a pursuit tracking task where the target intermittently and briefly disappears. You could then compare two versions of the PCT model in terms of how well they are able top account for the behavior in this experiment. One version would be the plain vanilla, sensory based perception version of the PCT model; the other would be a version that fills in the imagined position of the target when it disappears. If the model with the imagination component works best then you would have made a significant improvement to the PCT model.

RM: Anyway, just some thoughts derived from this thread. I’d love to hear your (and anyone else’s) thoughts about this.

Best regards

Rick

–
Richard S. Marken, Ph.D.
Author of Doing Research on Purpose.
Now available from Amazon or Barnes & Noble

In nature there’s no blemish but the mind

None can be called deformed but the unkind.

Shakespeare, Twelfth Night

          RM: Yes you should. I shouldn't have said what I said

and I humbly apologize. Not only was it wrong – you have
done much excellent empirical research – but it was an ad
hominum attack. I have no idea what got into me; I hope
you will accept my apology. I was wrong, wrong, wrong.

[MartinT]

[Martin Taylor 2014.11.29.13.59]

[From Rick Marken (2014.11.29.1030)]

Remember that the single "wires" in the diagrams in B:CP and

elsewhere represent conduits for “neural currents” that consist of
averaged firing rates in a large number of fibres. So although Bill
showed a switch, the actual situation is more likely to be a mix of
imagination and sensory data, biased one way or the other according
to the momentary situation.

It wouldn't be uncontrolled if the weights to the outputs depended

on the proportion of imagination in the input from that source to
the perception being controlled. In other words, if the last sensed
value of X comes from imagination, then the output would influence
only Y, just as would happen if X were firmly fixed in the
environment. Bill’s imagination-connection switch disconnects output
to X, and some such condition must happen in the biological system.
Even if X were severely disturbed, variation in Y could keep the sum
controlled. What won’t work is varying a reference value for the
X-level control unit when X has an imagined value.

Let Q = aX + bY when X is sensed, and

Q' = aX' + bY when X' is the imagined value of X (a constant k equal

to the last sensed value of X – or possibly a predicted value based
on X’s most recent velocity and acceleration, but let’s just say a
constant)

The question is what happens when Q' deviates from its reference

value R. It’s actually the same as any single-loop control case.
Varying the output “oy” to Y counters the disturbance “dy” to Y.
Meanwhile the disturbance “dx” to X remains unopposed, so the
environmental Q = aX+bY changes, but the perceptual Q’ = aX’ + bY
remains controlled.

That's one of the possibilities you can run with the general tracker

software. I’ve done lots of tracks under those conditions, but since
the general tracker doesn’t do modelling, I have not tried to do
anything with them other than just look at the traces. If you want
to try, I think that the version at
https://dl.dropboxusercontent.com/u/26441754/FlexibleTracker_b2d.zip
might still be functional, though Java has moved on since it was
written. I posted it in October 2013.

I don't think I would have changed the model in any way from what

Bill described, other than by mentioning that real nervous systems
don’t use neural currents conveyed by single wires. And Bill
described that, too, though the models and diagrams show and use
single wires.

Martin

···

            Martin Taylor

(2014.11.28.11.48)–

            MT: Very handsome of you, and I thank you for it,

while adding my own apology for participating in the
conflict escalation.

            MT: With luck, we should be able to terminate this and

the ancestor thread at this point.

          RM: OK, then why don't we start a new thread based on

your proposal about controlling perceptions partially in
imagination. In PCT as it now stands control is either of
sensory based perceptions – perceptions of aspects of the
environment – or control of imagined perceptions –
perceptions that are replays of reference signals
retrieved from memory (as per Figure 15.3, p. 223 of B:CP,
2nd Ed). So in the original version of PCT, whether we are
controlling perceptions or imagined perceptions was
conceived of as an all or none process – the result of
throwing the neural “switches” from one connection to
another.

                        RM: Yes you should. I shouldn't have said

what I said and I humbly apologize. Not only
was it wrong – you have done much
excellent empirical research – but it was
an ad hominum attack. I have no idea what
got into me; I hope you will accept my
apology. I was wrong, wrong, wrong.

          RM: You have proposed (and I know that we have

discussed this on CSGNet before so I know that Bill
agreed) that it is possible that the perceptual signals
that are controlled could be partially the result of
sensory based and partially the result of memory based
inputs to a perceptual function. The memory inputs would
be there to “fill in” what might be missing from the
sensory based input (due to brief obstructions or
whatnot).

          RM: I think it would be a great contribution to PCT if

you could determine how (and, of course, whether) this
mixed sensory/memory perception concept should be
incorporated into the PCT model. I just ran a quick little
simulation where I had systems controlling a sensory based
perception p = aX+bY and switched X from being an
environmental variable to being a reference signal from
memory(an imagined input, call it X’) and what I found
was that control of the now partially imagined perception
– p = aX’+bY – remained intact but the environmental
correlate of that perception – aX+bY seemed to be
completely uncontrolled.

          I haven't done the quantitative calculations yet to

determined whether control of aX+bY is really lost
completely when only X is imagined, so it may be true that
control of aX+bY is not completely lost when one component
of the perception of that variable is imagined. But this
would be a nice thing to determine to see how control of a
partially imagined perception might work.

          RM: Another aspect of this that might be good to look

into is whether having an imagined component of a
perception improves control when the sensory component is
lost, at least for a short time. I think you have
suggested that this may be the case and I think it’s is
possibly true. For example, suppose you have a person do
a pursuit tracking task where the target intermittently
and briefly disappears.

          You could then compare two versions of the PCT model in

terms of how well they are able to account for the
behavior in this experiment. One version would be the
plain vanilla, sensory based perception version of the PCT
model; the other would be a version that fills in the
imagined position of the target when it disappears. If the
model with the imagination component works best then you
would have made a significant improvement to the PCT
model.

[rsmarken]

[From Rick Marken (2014.11.30.1710)]

···

Martin Taylor (2014.11.29.13.59)

MT: Remember that the single "wires" in the diagrams in B:CP and

elsewhere represent conduits for “neural currents” that consist of
averaged firing rates in a large number of fibres. So although Bill
showed a switch, the actual situation is more likely to be a mix of
imagination and sensory data, biased one way or the other according
to the momentary situation.

RM: Yes, and that should be explicitly incorporated into your model. Bill was very exact in his modeling. If he had wanted to have the perceptual signal get inputs from imagination and sensory data simultaneously he would have shown that in Figure 15.3. I know that Bill thought that that kind of mixing occurred but, as far as I know, he never built a model that controlled the mixed signal. So if you developed a control model that controlled a percpetion that was a mix of imagination and sensory input that would be a very interesting extension of PCT. Whether it’s a necessary extension is another question, one that would have to be tested though experimentation.

MT: That's one of the possibilities you can run with the general tracker

software. I’ve done lots of tracks under those conditions, but since
the general tracker doesn’t do modelling, I have not tried to do
anything with them other than just look at the traces. If you want
to try, I think that the version at
https://dl.dropboxusercontent.com/u/26441754/FlexibleTracker_b2d.zip
might still be functional, though Java has moved on since it was
written. I posted it in October 2013.

RM: OK, I might give it a try. But I wish you would do this since it would help us determine whether this imagination/sensory input mix is really happening. I happen to think it’ s not but I would love to be convinced that it’s for real. You seem to think it is for real so it would be nice if you did the research and modeling to convince me that it is.

MT: I don't think I would have changed the model in any way from what

Bill described, other than by mentioning that real nervous systems
don’t use neural currents conveyed by single wires. And Bill
described that, too, though the models and diagrams show and use
single wires.

RM: I think you would have to explicitly put in the mixing function. And if you apply the model to the tracking task with intermittent disappearance of the target you would have to incorporate into the model the mechanism that switches in and out of imagination (if that’s what you think happens).I think a working computer model would really help us understand what is being proposed regarding the control of a mix of sensory-based and imagined perception. I think it would be a tremendously useful effort. I wish you would do it. You’re a very good programmer; you could probably knock out a working model in no time.

Best regards

Rick

–
Richard S. Marken, Ph.D.
Author of Doing Research on Purpose.
Now available from Amazon or Barnes & Noble

In nature there’s no blemish but the mind

None can be called deformed but the unkind.

Shakespeare, Twelfth Night

          RM: OK, then why don't we start a new thread based on

your proposal about controlling perceptions partially in
imagination. In PCT as it now stands control is either of
sensory based perceptions – perceptions of aspects of the
environment – or control of imagined perceptions –
perceptions that are replays of reference signals
retrieved from memory (as per Figure 15.3, p. 223 of B:CP,
2nd Ed). So in the original version of PCT, whether we are
controlling perceptions or imagined perceptions was
conceived of as an all or none process – the result of
throwing the neural “switches” from one connection to
another.

          RM: Another aspect of this that might be good to look

into is whether having an imagined component of a
perception improves control when the sensory component is
lost, at least for a short time. I think you have
suggested that this may be the case and I think it’s is
possibly true. For example, suppose you have a person do
a pursuit tracking task where the target intermittently
and briefly disappears.

          RM: You could then compare two versions of the PCT model in

terms of how well they are able to account for the
behavior in this experiment.

[wmansell]

Hi Rick and Martin, this certainly sounds like a potentially very important project to me! I hope may be shoe to help in some way.

Warren

···

Martin Taylor (2014.11.29.13.59)

MT: Remember that the single "wires" in the diagrams in B:CP and

elsewhere represent conduits for “neural currents” that consist of
averaged firing rates in a large number of fibres. So although Bill
showed a switch, the actual situation is more likely to be a mix of
imagination and sensory data, biased one way or the other according
to the momentary situation.

RM: Yes, and that should be explicitly incorporated into your model. Bill was very exact in his modeling. If he had wanted to have the perceptual signal get inputs from imagination and sensory data simultaneously he would have shown that in Figure 15.3. I know that Bill thought that that kind of mixing occurred but, as far as I know, he never built a model that controlled the mixed signal. So if you developed a control model that controlled a percpetion that was a mix of imagination and sensory input that would be a very interesting extension of PCT. Whether it’s a necessary extension is another question, one that would have to be tested though experimentation.

MT: That's one of the possibilities you can run with the general tracker

software. I’ve done lots of tracks under those conditions, but since
the general tracker doesn’t do modelling, I have not tried to do
anything with them other than just look at the traces. If you want
to try, I think that the version at
https://dl.dropboxusercontent.com/u/26441754/FlexibleTracker_b2d.zip
might still be functional, though Java has moved on since it was
written. I posted it in October 2013.

RM: OK, I might give it a try. But I wish you would do this since it would help us determine whether this imagination/sensory input mix is really happening. I happen to think it’ s not but I would love to be convinced that it’s for real. You seem to think it is for real so it would be nice if you did the research and modeling to convince me that it is.

MT: I don't think I would have changed the model in any way from what

Bill described, other than by mentioning that real nervous systems
don’t use neural currents conveyed by single wires. And Bill
described that, too, though the models and diagrams show and use
single wires.

RM: I think you would have to explicitly put in the mixing function. And if you apply the model to the tracking task with intermittent disappearance of the target you would have to incorporate into the model the mechanism that switches in and out of imagination (if that’s what you think happens).I think a working computer model would really help us understand what is being proposed regarding the control of a mix of sensory-based and imagined perception. I think it would be a tremendously useful effort. I wish you would do it. You’re a very good programmer; you could probably knock out a working model in no time.

Best regards

Rick

–
Richard S. Marken, Ph.D.
Author of Doing Research on Purpose.
Now available from Amazon or Barnes & Noble

In nature there’s no blemish but the mind

None can be called deformed but the unkind.

Shakespeare, Twelfth Night

          RM: OK, then why don't we start a new thread based on

your proposal about controlling perceptions partially in
imagination. In PCT as it now stands control is either of
sensory based perceptions – perceptions of aspects of the
environment – or control of imagined perceptions –
perceptions that are replays of reference signals
retrieved from memory (as per Figure 15.3, p. 223 of B:CP,
2nd Ed). So in the original version of PCT, whether we are
controlling perceptions or imagined perceptions was
conceived of as an all or none process – the result of
throwing the neural “switches” from one connection to
another.

          RM: Another aspect of this that might be good to look

into is whether having an imagined component of a
perception improves control when the sensory component is
lost, at least for a short time. I think you have
suggested that this may be the case and I think it’s is
possibly true. For example, suppose you have a person do
a pursuit tracking task where the target intermittently
and briefly disappears.

          RM: You could then compare two versions of the PCT model in

terms of how well they are able to account for the
behavior in this experiment.

[MartinT]

[Martin Taylor 2014.12.01.09.56]

[From Rick Marken (2014.11.30.1710)]

I have no idea whether mixing of the kind I suggested ever happens.

My guess is not at low levels, but yes at higher levels. My main
point is that if you remember that the single wire approximation in
Figure 15.3 and every other PCT figure actually represents a myriad
of fibres, not all of which come from the same place or go to the
same place (neurons synapse on thousands of other neurons), then if
the “wire approximation” of Fig 15.3 has any relation to the truth,
there is no need for a pair of switches. Flipflop connections using
lateral inhibition could serve as both switches and mixers,
depending on the strengths of the cross-connections. It’s not an
either-or proposition. (And I’m not asserting that flip-flop
connections exist in those parts of the circuit, either; I’m just
noting that if you remember that the “wire” is only an
approximation, then these possibilities exist and are entirely
compatible with Bill’s writing).

No, I don't think it is real. Nor do I think it is not real. Nor do

I want to convince you of anything. I think it is possible, and
given that the “wire connection” represents a bundle of fibres,
there’s a plausible mechanism for it to happen. That’s all.

What I do think is real is that in a perception built from many

components, it is often true, at least at the higher levels, that
you don’t have incoming sensory data on some of the components,
which must then come from memory. How much of your perception of the
present political state of the USA is based on currently incoming
sensory data? Closer to home, I imagine you control a perception of
the roadworthiness of your car, and that you have a current
perception of how roadworthy it is. How much of that is based on
sensory data you are acquiring at this moment? How much while you
are on the road? How much when you are discussing it with a mechanic
after an inspection?

If I were to try something of that kind, it would be a long time in

the future. Right now, I have a lot of other stuff on my plate. I
realize Warren thinks it important, but I’m not sure why. To me it
seems clear that we do use imagination/memory along with sensory
data to form our perceptions. Whether imagination and incoming
sensory data mix at the output of any particular perceptual function
is interesting, but I wouldn’t know how to set up an experiment that
could distinguish between models in which that happened and models
in which the imagination switched in or out, all or nothing, at a
lower level.

Martin

···

            Martin Taylor

(2014.11.29.13.59)

            MT: Remember that the single "wires" in the

diagrams in B:CP and elsewhere represent conduits for
“neural currents” that consist of averaged firing rates
in a large number of fibres. So although Bill showed a
switch, the actual situation is more likely to be a mix
of imagination and sensory data, biased one way or the
other according to the momentary situation.

          RM: Yes, and that should be explicitly incorporated

into your model. Bill was very exact in his modeling. If
he had wanted to have the perceptual signal get inputs
from imagination and sensory data simultaneously he would
have shown that in Figure 15.3. I know that Bill thought
that that kind of mixing occurred but, as far as I know,
he never built a model that controlled the mixed signal.
So if you developed a control model that controlled a
percpetion that was a mix of imagination and sensory input
that would be a very interesting extension of PCT. Whether
it’s a necessary extension is another question, one that
would have to be tested though experimentation.

                        RM: OK, then why don't we start a new

thread based on your proposal about
controlling perceptions partially in
imagination. In PCT as it now stands control
is either of sensory based perceptions –
perceptions of aspects of the environment –
or control of imagined perceptions –
perceptions that are replays of reference
signals retrieved from memory (as per Figure
15.3, p. 223 of B:CP, 2nd Ed). So in the
original version of PCT, whether we are
controlling perceptions or imagined
perceptions was conceived of as an all or
none process – the result of throwing the
neural “switches” from one connection to
another.

                            MT: That's one of the possibilities you can run

with the general tracker software. I’ve done lots of
tracks under those conditions, but since the general
tracker doesn’t do modelling, I have not tried to do
anything with them other than just look at the traces.
If you want to try, I think that the version at https://dl.dropboxusercontent.com/u/26441754/FlexibleTracker_b2d.zip
might still be functional, though Java has moved on
since it was written. I posted it in October 2013.

          RM: OK, I might give it a try. But I wish you would do

this since it would help us determine whether this
imagination/sensory input mix is really happening. I
happen to think it’ s not but I would love to be convinced
that it’s for real. You seem to think it is for real so it
would be nice if you did the research and modeling to
convince me that it is.

                        RM: Another aspect of this that might be

good to look into is whether having an
imagined component of a perception improves
control when the sensory component is lost,
at least for a short time. I think you have
suggested that this may be the case and I
think it’s is possibly true. For example,
suppose you have a person do a pursuit
tracking task where the target
intermittently and briefly disappears.

            MT: I don't think I would have changed the model in any

way from what Bill described, other than by mentioning
that real nervous systems don’t use neural currents
conveyed by single wires. And Bill described that, too,
though the models and diagrams show and use single
wires.

          RM: I think you would have to explicitly put in the

mixing function. And if you apply the model to the
tracking task with intermittent disappearance of the
target you would have to incorporate into the model the
mechanism that switches in and out of imagination (if
that’s what you think happens).I think a working computer
model would really help us understand what is being
proposed regarding the control of a mix of sensory-based
and imagined perception. I think it would be a
tremendously useful effort. I wish you would do it. You’re
a very good programmer; you could probably knock out a
working model in no time.

                      RM: You could then compare two versions of

the PCT model in terms of how well they are
able to account for the behavior in this
experiment.