RKC-RICK'S DECSNS

<Bob Clark (940628.1435 EDT)>

Tom Bourbon [940621.0943] Subject: Re: RKC-PROGRAM CONTROL

Bob,

...

Maybe I should use one of Martin's "alerting stimuli" on you. That
would force you to pay attention wouldn't it? ;-))

ASIDE

Calling someone's name is one of the standard ways for attracting
attention. "Force" seems a bit too strong, implying a loss of
control by the one called.

While on the subject, briefly, I hope, "attention" can be "attracted"
in many ways by many events. It seems to me that any unexpected
combination of sensory signals will get attention, but usually only
briefly while it is being evaluated for significance. The concept of
an "alerting system" seems to me quite unnecessary, although
meaningful for one desiring to "alert" someone else. Otherwise, such
events are readily described and understood without any special
"alerting system."

END OF ASIDE

program control deserves attention on its own. Twice, I've tried to
attract your attention on that subject,

At the time of your current post [Tom Bourbon [940621.0943] Subject:
Re: RKC-PROGRAM CONTROL] I had not sent my response of 06-21-94. Now
that you are back at home, you will have seen it.

I saw your previous post [Tom Bourbon [940603.1108], but it didn't
seem to call for comments from me. I no longer have a copy available.

Your more recent post, 940616.1640, included a summary of your
earlier post, so I am looking at that version.

In this summary several items are unclear to me (I am sure that they
are very clear to you).

I adopted a plan to create a sequence of four different
relationships between the cursor and target during a single
one-minute run of a pursuit tracking tracking task.

Does this mean that the computer generated a more-or-less standard
pattern that could be tracked with the cursor controlled by the
experimental subject? I think so.

And that the subject selects which of the "four relationships" he
will display on the monitor? And when they are to be displayed?

That's at least three different levels of control in the same run.

I'm afraid I don't know what you mean here by "different levels of
control."

I decided that a change from one controlled relationship to the next
should occur every 15 seconds.

A signal ("beep") was used to tell the subject when he should change
the specific relationship he had been instructed to maintain.

You next described the four relationships the subject is to use. ok.

The run began, the "signals" occurred

and the subject followed the instructions.

I do not believe I used a "program" of movements. Instead, I used a
program of time-dependent changes in my intended perceptions of
relationships between cursor and target.

To your first sentence:
It seems to me that moving the cursor to achieve any desired
perception of cursor position requires that you have systems
corresponding to the connections between the handle and the cursor.
This might be termed a "program" of movements. But for the purposes
of this demonstration, you need not point out this relationship.
Surely some of your sub-systems are used in moving the handle to
place the cursor where you wanted to perceive it. Although both the
handle and the cursor must have been moving, briefly, your attention
was on the final position, not the movement in getting there. ok.

As to the next sentence, "Instead etc," I have no idea what you mean.
I am unable to even guess what you may have in mind.

Next you

did a second run, in which a random disturbance, never before
encountered by me, also affected the position of the cursor.

I am immediately curious as to the nature of this "disturbance"
whence it came -- was it in the computer? was there another
participating party? did it affect the connection between the handle
and the cursor? etc?

Also, in what way was it random? in its timing? in the nature of its
effects? etc?

And how did you achieve the property of being "random?"

I have included other comments in my recent post: Bob Clark
(940621.2125) Subject: Re: Re: RKC-PRO CONT.

I hope these comments are useful.

Regards, Bob Clark

[Martin Taylor 940628 16:40]

Bob Clark (940628.1435 EDT)

I'm off in a few hours until around July 11. See you later.

While on the subject, briefly, I hope, "attention" can be "attracted"
in many ways by many events. It seems to me that any unexpected
combination of sensory signals will get attention, but usually only
briefly while it is being evaluated for significance. The concept of
an "alerting system" seems to me quite unnecessary, although
meaningful for one desiring to "alert" someone else. Otherwise, such
events are readily described and understood without any special
"alerting system."

I'm not clear what "unexpected" means to an ECS in the standard hierarchy.
I'm even less clear how a control system that deals in the level of a
single scalar valued "perceptual signal" can treat an "unexpected
combination of sensory signals." I become completely foggy on how
an ECS as normally considered can "evaluate for significance" such an
"unexpected combination of sensory signals."

It seems to me that you have in mind some very major alteration to the
structure of HPCT, beyond what you already propose with the DME. I don't
understand what that alteration might involve, but it is clearly a much
greater alteration than is implied by the construct of "alerting," which
may not need any change in the structure at all. It remains to be seen
whether the classical structure can handle the alerting phenomenon. The
"alerting system" is not necessarily any special hardware, though it may
turn out to involve special hardware (as in the visual periphery, which
seems to provide the clearest physiological example of a system developed
primarily for alerting.

It is incumbent on anyone proposing additional complexity to a working
model that they show where the addition improves the range or the precision
of the model's application. I am not sure that you have shown the need
for specialized subsystems that look for "unexpected combinations of
sensory events." If you think they are necessary, you should (a) explain
how they might work--it's a problem looking for the unexpected--and (b)
show how they improve on the "standard model."

In my view of alerting, only anticipated combinations of input data
are involved, such as the combinations that occur in any perceptual input
function. The question of alerting is whether the level of an EXPECTED
combination has suddenly changed, or has exceeded some reference level.
The alerting phenomenon may or may not be a normal performance of the
classic hierarchy, but whether it is or not, I do not think it requires
new kinds of subsystem.

Martin