Applications--HCI

[Martin Taylor 950817 12:00]

Rick Marken (950816.2135) to Bill Powers

Back again, with 300 e-mail items piled up--but this one is today's.

Maybe it would help people understand the PCT model of behavior if you could
describe your vision of an application in which PCT is used as the
main organizing principle.

I thought you did so very well yourself in >>Rick Marken (950815.2130)

Actually, I have posted a little about my use of PCT in my work. I think
the most systematic application of PCT was to the analysis of satellite
ground control operations. This "task analysis" was done to provide a
framework for the design of human-computer interface (HCI) software
for integrated (multiple) satellite control.

The PCT based task analysis was particularly useful for HCI design
because it centered around the specification of the variables that the
operator had to control (which would be the variables that had to be
displayed to the operator in some way) rather than on the (in practice,
unpredictable) sequence of actions that the operator should take to control
those variables (the typical approach to task analysis). The PCT based
task analysis also specified the disturbances that might affect the variables
to be controlled (component failures, electromagnetic interference,
scheduling conflicts, etc) and the actions (computer inputs from the
operator) that the operator would have to be able to take to counter
these disturbances.

Maybe I like this so much because it is exactly the basis for my Layered
Protocol design tool for interfaces. And it is what I hoped you would say
in the paper you have proposed for and withdrawn from the IJHCS special
issue on PCT. Such a paper would provide a great centre-piece for the
issue.

I think that the approach to task analysis you describe is the only one
that can reliably lead to "good" interfaces (i.e. ones with which operators
feel comfortable and that unobtrusively lead them to act appropriately to
bring the perceptions under control that are required for the system to
produce the results desired by the ultimate users--who are not usually the
operators). All task analysis procedures I know deal with how the operator
must act in order to permit some other action. But I think I am managing
to get across the PCT approach you describe to at least one, maybe two
of our human engineers. Slowly, but slowly...

Anyway, good stuff.

This task analysis was done with the help of satellite controllers who
liked the whole approach very much; there was no resistance to this PCT
analysis because it didn't conflict with anyone's existing paradigms;
these were satellite controllers and they knew that their job was to
control variables represented on the computer display.

Better yet!

I think most people in my field (HCI engineers) work from a PCT
perspective whether they call it that or not (mostly not).

But I don't agree with this, any more than you agree that most psychologists
understand control, whether they call it that or not. I think most HCI
people deal with the action side of the interface, providing the right
controls, and often forget the displays that show the physical or internal
variables for which controllable perceptions should be available to the
operators. In other words, they tend to provide means to affect the CEVs
that are important to the user, without providing the means for the
operators to perceive easily the states of those CEVs.

(Parenthetical example: I was at my bank yesterday to sort out an overdraft
that I didn't think should have occurred. The tellers had signs to the
effect that customers should expect delays because they were learning a
new computer system--Windows-based, as one might expect from the sign.
When I posed my problem to the teller who I have previously found to be
the most astute, she started going through the menus, trying several times
to bring up my recent transaction record. Eventually she had to ask the
trainer who was sitting behind. The trainer asked what kind of account
it was--a personal chequing account, the most common kind--and then said
that it was impossible to bring up the transaction record for that kind
of account. The teller had to quit the Windows system, type something like
PCAST (which she wrote down in a little notebook) and then use the older
system, with which she was familiar. I would have thought that the
perception of the state and history of various accounts would have been
one of the primary things that the designers of the new system would have
provided for the tellers. But no. They provide lots of menus that allow
for all sorts of actions, some of which SEEM as if they should lead to the
desired result, but don't. In the end, we found a withdrawal that
shouldn't have been there, but even with the old system she couldn't tell
who made it, and for that I had to go to a different desk. The answer took
another 6 hours to get to me, as somebody downtown had to use yet another
system--probably paper-based--to find out.)

Anyway, HCI seems a very fertile field of applications for PCT-based
discussion, and I think you started it very well.

Martin