[
Lists Home |
Date Index |
Thread Index
]
I agree about the expense, but in studying the web as a
system, it is useful to forgo the technical web architecture
and come to understand that its complexity emerges from
the interaction of agents both human and automated, and
for the semantic web future, agents empowered to act
on behalf of human owners, thus The Golem Problem I
discussed here and in the Markup Languages MIT magazine
article.
Asking isn't the whole solution, but even in agent-based
modeling, it is an important task. For many
kinds of conflicts (say the human profile), it is
the only appropriate place to start. The toughest
problem of agent-based modeling is not knowing
what the human owner wants, and worse, that the
human does not know what they want. They act
on habit formation which might be called 'instinct'.
Understanding habit formation with regards to
symbol/sign assignment is important to finding
the conflict of semantics. The interface aspects
are VERY important and why marketing designers
and for their own good, web GUI designers, should
study semiotics. And so should we because semiotics
are important to how and why web systems become complex.
Here are some notes from the article you reference for
others who are only lightly following this thread, to
illuminate the topics, edited for brevity so not
exactly quoted. Comments in brackets are mine. I start
with the Eurobios site definitions:
Also from Eurobios site:
Complex Systems: elaborate and unpredictable properties arise from
interacting
agents. Examples of emergent properties include how the system organizes
itself,
how it finds balance of order and disorder, how agents individually and
collectively
evolve new behaviors in response to change.
Emergent behavior: emerges from interactions with other agents and
environment
rather than being imposed because behavior cannot be deduced from the rules.
Co-adaptation: actions of all agents in environment affect each other, such
that they co-adapt and co-evolve. Competitive advantage is gained by
effectively
adapting to novel and unpredictable situations faster than the competition
(seeing
the setup).
Then the Economist article:
"Complex Agent-Based Dynamic Network: explain behaviour through the use of
agents:
a program that acts in a self-interested manner in its dealings with
numerous other
agents inside a computer. Can mimic almost any interactive system. If the
constituent
parts can be understood, some insight into the whole will follow.
[The goal is prediction. Simulation or role play outs the intent of the
communication: what the agent wants. Hint: what are the public and
private couplers?]
One of the challenges in setting up an agent-based model is defining clearly
what individual agents want.
Not only are people often inaccurate in their beliefs about themselves,
but especially in the business world they lie.
[Measure]: she set out to measure what was actually going on. Spent hours
watching and recording which
machines were running when. [See Fisher Information]
[Entropy as measure of system disorganization]
Compared her findings about what was really happening with what people
claimed
was happening on paperwork such as invoices. By checking how often these
agreed,
she could approximate the mathematical entropy of the system: a measure of
how disordered
it was.
[Entropy as measure of real time knowledge. Compare Boltzman and Shannon
entropies and the concept of addressing. Why URIs and the problem of
URI management]
A complicated factory, with many different assembly lines, can still be
ordered
if everything is proceeding according to plan. Even a simple one, by
contrast,
would be disordered if managers had no idea of what was happening."
len
From: Didier PH Martin [mailto:martind@netfolder.com]
Practically, since we tried that in the past, asking to humans to resolve
issues is long and costly process. Moreover, I may have some difficulties to
ask the others to resolve all ontological conflicts I will encounter on the
web. However, I think, intuitively, you point to a good direction. More and
more I think that agents based modeling can help a lot for this type of
phenomenon hard to model the usual way.
For those of us who want to explore the agents based modeling here is a good
link. http://www.econ.iastate.edu/tesfatsi/ace.htm. It's about using agents
based modeling in the context of economics but it could be very well applied
on the topic of concept categorization or concept clustering. The site links
to a lot of articles with a pedagogical perspective.
I think that agents based modeling can be tremendously useful for a
_pratical_ semantic web.
|
