Roger -
I think this is a good analysis of Russ's idea/question/pursuit.
But/And that is what makes it an interesting pursuit. Self-modifying
code has always been a hairy-scary mess but/and we have slowly found
ways to tame it enough to make it useful in various situations.
Two platitudes I've heard coming out of complexity science :
1) Constraint provides form
2) When stuck solving a problem, add another level of indirection
The first platitude refers to why there might not be many/any ABM
systems which offer self-modification (as Roger alludes to here)... so
far, nobody has been able to release the constraint and still get
enough coherence out to be useful/interesting.
The second platitude refers to why one might *want to* do what Russ
wants to. It would seem that this form of self-modification is a new
level of indirection.
I wonder if the combined genius, memory, attention, focus of this group
can come up with real-world systems which (obviously) seem to work this
way (require this level of abstraction/complexity to model).
-Steve
I don't think you'll find this because it implies
programming a higher purpose and allowing the agents to jump the rails,
as it were, and start negotiating their way through the combinatorics
of alternative networks. Similarly, you won't find models in which
agents invent new inputs to monitor, new outputs to generate, and new
rules which involve new inputs and new outputs.
Optimization within a fixed solution space, which is what we do when we
let agents play with the flow through a fixed network or let them
search out the most profitable rules in a set of prespecified
alternatives, gets hairy enough without opening things up to the
infinity of potential solutions that we didn't have time to program
into the model ourselves.
Neal Stephenson's novel Anathem has some discussions, central to the
plot, about how the human mind filters the combinatorically possible
down to the mechanically feasible and further down to the set of
outcomes worth thinking about.
It's not obvious how to program the same capabilities without the
teaching the agents how to apply the same common sense and expert
senses which the programmer uses to frame a fixed solution space.
-- rec --
On Fri, Aug 28, 2009 at 6:48 AM, Russ Abbott
<[email protected]>
wrote:
In
a discussion with a colleague today we talked briefly about stocks and
flows networks. It struck me that a stocks and flows model is a limited
sort of service-oriented agent-based model. In a service-oriented
agent-based model, agents accept inputs and produce outputs -- the
simplest version being a supply chain. That's really a stocks and flows
model in which the agents control the flows. Important differences are:
- In an agent-based model, the agents are assumed to be
autonomous in various ways. In a stocks and flows model the flow rates
are not autonomous. The flow rates are equations that don't have the
ability to change themselves.They are assumed to be facts about the
nature of the domain being modeled.
- In a service-oriented agent-based model the agents have the
ability to reconfigure themselves dynamically and perhaps even to add
new agents and new stock nodes. In a stocks and flows model, the
structure of the network static.
So this raises the question whether anyone knows of any work in stocks
and flows modeling that addresses stocks and flows networks that are
flexible in the ways just mentioned.
-- Russ
============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
lectures, archives, unsubscribe, maps at http://www.friam.org
============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
lectures, archives, unsubscribe, maps at http://www.friam.org
|
============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
lectures, archives, unsubscribe, maps at http://www.friam.org
