Lists Home |
Date Index |
The topic you may wish to research is the application
of analogical reasoning systems. Before we get into
this deeply in this thread, I suggest a review of
Some background in semiotics (sign processors or
a 'species independent theory of cognition, of
simply, learning from experience) is useful but not
essential. The essential position is:
We may deduce a choice, but to get the initial
lists of choices to choose from, we typically use
analogical means. So we proceed from abduction,
to induction, and then to deduction.
The issues of drunkenness, menses, current emotional
state, etc., have some potential solutions in analogical
systems but as I said, fall into the category of
Quality of Service.
We may wish to dig more deeply into
types of analogical engines, the operations by which
analogies and analogies of analogies are created, and
the probability or weight of evidence.
Not so oddly, given XSLT selectors over XML documents,
Sowa's methods might be implementable without resorting
to exotic technologies. That's a hunch, not an established
fact. There are operations and rules specified in the
Sowa paper, and it would be interesting if the XSLT
gurus took a look at those and commented on how many
and with what level of performance they could be
implemented in XSLT.
From: Linda Grimaldi [mailto:firstname.lastname@example.org]
OK, after my rant on automated voice systems, let me play devil's advocate
for just a moment- not sure my heart's in it, but I'll give it a try.
I have gotten equally, if not more, frustrated with human customer support
interactions- try calling your medical insurance provider one day for a
virtually guaranteed example.
While we carry on about the risks of allowing a computer to make a decision,
human judgement is also inherently flawed (look at the decision to go into
Iraq- oops- just got political).
And it's not necessarily due to lack of good information. The nice thing
about computers is that they have no axes to grind other than the ones they
have been programmed with, don't suffer from PMS or other emotional
disorders and aren't in a bad mood from dealing with the kids and a
checkbook that is always in the red. To use Steve's example, if I am coming
home drunk and disorderly, my judgement just may fail me at the critical
moment when I turn on that light myself.
So, my question is (and it's not entirely tongue in cheek): can automated
computer service systems ever hope to provide a level of service that,
while not perfect, is at least in some respects the equal of that provided
by the average human customer service interaction? And what kinds of
service scenarios are best left to humans, and what kinds to computers? I
don't necessarily think that only the plebes will get XML automated
services, while those who pay will get humans. I think it will depend on
the service type- and the plebes may get the humans who hear voices...