# Delay in human controller

**URL:** <http://discourse.iapct.org/t/delay-in-human-controller/3405>\
**Category:** CSG1996\
**Created:** [September 10, 1996, 9:46pm UTC](http://discourse.iapct.org/t/delay-in-human-controller/3405 "1996-09-10T21:46:18Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![Bill\_Powers1](http://discourse.iapct.org/letter_avatar_proxy/v4/letter/b/8e7dd6/32.png) [@Bill\_Powers1](http://discourse.iapct.org/u/Bill_Powers1)\
**Post date:** [September 10, 1996, 9:46pm UTC](http://discourse.iapct.org/t/delay-in-human-controller/3405/1 "1996-09-10T21:46:18Z")

</div>

[From Bill Powers (960910.1415 MDT)]  
Hans Blom, 960910d--

> The human controller has a built-in time delay of about 0.5 seconds, I  
> guess, the MCT controller doesn't. This can be tested: to a human, offer

the \>feedback information at, say, 1 second intervals; this would make the delay

> time negligible to the time within which a reaction is required.

In the best-fitting model of the human controller, we have an integral  
output function and a transport lag (put into the perceptual function). The  
best fitting transport lag for most people is about 0.16 seconds, or about  
10 frames of the display at 60 frames per second. If you gave a human being  
feedback information only once per second, tracking would be almost  
impossible, using a disturbance with a bandwidth of about 0.25 Hz as we do.

When Rick did the reversal experiments, I believe the sampling rate (for  
both human and model) was 25 per second -- the Mac has a slower frame rate  
than a PC. The exponential runaway seen in both human and model, which  
lasted about 0.4 seconds after the reversal, thus was sampled about 10  
times, giving the opportunity for a good fit of an exponential to the data.  
It was clear that the control system in the person did not reverse its sign  
for 0.4 seconds, and that the consequence was the beginning of an  
exponential runaway due to positive feedback.

Note that a transport lag of 0.16 sec does not mean that the behavior is a  
series of corrections 0.16 sec apart. The output is continuously variable;  
it is simply delayed by 160 milliseconds behind the continuous input  
variations. In the past I tested the effect of sampling rate up to 1000 per  
second, to make sure that a rate of 60 per second gave a true picture of the  
input-output function. At 60 per second I got the same results as at 1000  
per second. The 25 per second rate Rick used, forced by using a Mac, does  
begin to show some differences from the data obtained at higher rates, but  
they are small.

In armftau, the artificial cerebellum model using real dynamics, there are  
time-ticks on the display showing tenths of a second. The value of dt, or  
the time per iteration, is set, I believe, at 0.01 sec, and the physical  
dynamics of the arm are computed accordingly. The mass, length, and moment  
of inertia properties are those of my own left arm.  
................  
RE: Tom Bourbon's E. coli gain adaptation

> I hope that this makes clear why I say that (in a certain sense) a  
> controller must be the inverse of the world in which it lives. Even a  
> PCT controller...

In fact the output function in the model in question was an integrator, and  
the environmental feedback function was a proportional connection, but never  
mind. A integral might be said to be the inverse of a constant of  
proportionality in some cultures, although not, I suspect, among  
mathematicians. When you say "in a certain sense" I hope what you mean is  
"in an uncertain sense."  
...............  
Hans Blom, 960910g--

RE: perception

> It is the level of relating a measurement y (of something that  
> presumably exists in the real world) to some thing x that exists  
> internally (in the model).

You should really be calling that variable x', shouldn't you, as it's the  
output of the internal world-model? The one-to-one relationship I am talking  
about is between y, the perceptual signal, and the variable x which exists  
outside the sensors, in the environment. You treat y as if it is simply a  
measurement of a corresponding variable in the external world. The question  
I raised concerned perceptual signals y which are not direct functions of  
single external variables (or vectors), but which are functions of multiple  
external variables, with no one-to-one correspondence. For example, y might  
stand for the retinal distance between two objects in the environment. This  
distance must be computed from the relative positions of the two objects on  
the retina; there isn't any "thing" out there that corresponds to a  
distance. If p is a measure of position, then distance = p2 - p1. At least a  
subtraction is necessary to compute the distance. Of course there are many  
more complex examples of this.

> "A before B" relates two events, i.e. two things that occur at a  
> certain moment in time. If an equation reads, e.g.

> B (t) := f \* A (t-6\*dt)

> and all other coefficients are zero, we can say that changes in B  
> follow changes in A after a time 6\*dt and with a strength f.

> Does this clarify things?

Not much. You're saying that perception of "before" requires that there be  
someone who can look at the equation and interpret it to give a meaning of  
"before" to the relations described by the equation. But we can't have  
little people inside a model interpreting the equations; the model has to do  
that interpreting by itself. So you still haven't said how the model  
generates a perception of beforeness. You've only described the conditions  
necessary for such a perception to be appropriate.

P.S. I am now receiving only single copies of your post, via CSG.  
Congratulations.

> **···**
>
> --------------------------------------------------------------------------  
> Bruce Abbott (960910.1400 EST) --
> 
> > No, no, no, no, you miss the point. Simon is not making the Argument from  
> > Design. Natural selection does not involve a Natural Selector, i.e., Mother  
> > Nature or God, as you well know (and that fact necessitates the passive  
> > voice -- there is no one doing the selecting).
> 
> Simon didn't \_mean\_ to be arguing from design, but that is the form his  
> argument took. Can you fix up Simon's argument so the parallel holds true?  
> If there is a purpose for Y, then Y is employed to realize that purpose. In  
> the case where there is no agent X, there is nothing employing Y, and  
> nothing to have a purpose regarding the use of Y. How can you make Simon's  
> argument NOT come out to be an argument from design?
> 
> Our problem here is that a pun is involved. The word "purpose" refers to a  
> reference signal in a control system. It also refers to the (or rather, a)  
> use to which something could be put. These two usages involve completely  
> different meanings, as in the sentences "I saw the book" and "I saw the  
> plank." The use to which something could be put is not the same thing as a  
> reference signal. To confuse these two meanings is to court chaos, as when  
> you hear me say "I saw the plank" and don't know whether I'm talking about  
> vision or carpentry.
> 
> If I say "the purpose of the robot is to weld the piece of metal in its claw  
> to the stationary workpiece," you can't know what I'm talking about aslong  
> as we confuse these two meanings. Do I mean that inside the robot there is a  
> reference condition defined as "this piece of metal welded to that  
> workpiece," or do I mean "this robot was constructed to weld the two pieces  
> together, to satisfy the intentions of the production manager?" In other  
> words, am I referring to a purpose held by the robot and concerned with  
> certain parts of its environment, or am I referring to something in the  
> designer's or owner's mind concerned with the use to which the robot could  
> be put?
> 
> I think that the reason purpose gets involved in so much verbal trickery is  
> that there hasn't been an adequate definition of it prior to PCT. We're  
> watching people trying to figure out what purpose is without any model to  
> give it a clear definition. Even a Nobel Laureate can't mount a convincing  
> argument under those conditions, except by employing puns.
> 
> > The main purposes of a bird's wing are to provide the lift and thrust  
> > necessary for powered flight (although they serve other purposes as well).
> 
> That's a good example of what I mean. I am quite sure, considering the  
> knowledge we share, that you do not mean that there is a reference signal in  
> the bird's wing that specifies a purpose to be achieved. What you mean is  
> that if the bird intends to fly, the wing can be put to that use. If the  
> bird wants to splash water on itself in a birdbath, the wing can be put to  
> that use, too.
> 
> I claim that there is no use inherent in any object, in the absence of  
> organisms. The idea that purpose inheres in objects is simply indefensible.  
> No matter what purpose you claim an object has, it is always possible to  
> find a different purpose for the same object. That is because purposes are  
> in the uses of objects, not in the objects. What is the purpose of the Moon?  
> To cause tides? I'm sure that you would reject that idea. But how is it  
> different from saying that the purpose of a wing is to power flight?
> 
> Best,
> 
> Bill P.  
> ---------------------------------------------------------------------------