Re: [Swarm-Modelling] Re: [Swarm-Support] IDEs - who is using what

[glen e. p. ropella]

Darren Schreiber wrote:
> At a conference on "New Methods in Political Science" recently,  someone 
> was asking why ABM's still haven't had the payoff that was  promised a 
> decade ago in our field.  I believe this anarchic state  and a 
> continuing uncertainty about its epistemological foundations is  the 
> reason.

I agree.  However, I think it's unwise to push for premature 
convergence, unless the reasons for pushing for such convergence are 
based on some self-interested objective of obtaining and exploiting your 
own intellectual property, in which case, pushing for premature 
convergence can help you gather enough resources to embark on a 
large-scale effort -- public or private.  But, thinking altruistically 
(as a charitable organization like the SDG should), the anarchy is 
appropriate because the discipline is immature and its subject complex.

To profess or pursue something other than anarchy would be short-sighted 
if energetic anarchy is the best method we, as components of an 
algorithm, have for searching the space of our complex subject.

I.e. if there are no regularities in the epistemological foundations, or 
if we just haven't identified them yet, then pretending that there are 
or pretending any one point of view is now known to be right, will do 
more to limit the growth of the practice than the anarchy.

> [I'm hoping we can move this discussion to Swarm Modelling since  there 
> are people who monitor that list only who would likely have  insights.]

Thank you, especially since this thread is _really_ about the act of 
modeling, not support for the Swarm toolkit.  Back to Steve's and Chris' 
original questions, we're not _supposed_ to be "programming", per se. 
We're supposed to be modeling.  So, my answer to all of the above 
questions is: if a tool helps you model something, then it's a good tool.

Eclipse (which I use fairly often), as nice as it is as a programming 
tool, is all about the software.  It is a tool for specifying 
computational processes.  If and when it helps you "model", you're 
modeling the implementation, not the referent of a simulation.  Hence, 
it's not a good modeling tool for me.

FWIW, I have the same general problems with Swarm and Repast.  As good 
as they are in helping the user be computationally expressive, they're 
not doing a very good job at all of helping us build computational 
devices that _refer_ to other artifacts.

Mason and more streamlined tools like it are better, but only because 
they are simpler and require less investment in the computational 
infrastructure and allow you more time to think about the act of modeling.

Emacs (which I use most of the time) has the same problem as Eclipse. 
But, because emacs is more of a "shell" for the machine, it doesn't 
really pretend to be anything other than a computational tool.  It is 
sometimes touted as a "text editor" or a "text manipulation" tool (as we 
use it in Swarm, anyway).  But, one could say the same thing about the 
other "shells" we use.  It's an algebraic way of interacting with the 
machine.  This results in much less hubris than tools like eclipse about 
how useful it is to "program" using emacs.

Overall, we still can't beat Stella.  As modeling tools go, Stella takes 
the prize.

Having said that, I'd like to plug my own favorite IDE that has much 
more in common with Alessandro's (which is worth checking out if you 
actually want to model instead of program).  Ptolemy 
<http://ptolemy.eecs.berkeley.edu/> is a good modeling tool.  Because 
it's goal is modeling, it also takes care of many of the IDE type things 
a modeler needs to deal with.  I highly recommend Ptolemy.  Ptolemy II 
is, unfortunately, written in Java.  But, life is like that, eh?

> Gary Polhill wrote:
> >  I'd suggest not, because such a standardisation would impose the  
> > wrong kind of constraint on the way we go about building models. If  
> > we do standardise, I think it should be to constrain ourselves  
> > methodologically rather than technologically.

I'm a big fan of Paul Feyerabend's book, "Against Method" (but 
admittedly I like him mostly for his irreverent attitude ;-). 
Ultimately, I don't think constraining ourselves methodologically or 
technologically is necessary or wise.  As Gary stated in his post, the 
whole point of using agent-based techniques in the first place is to 
become more intimate with the mechanisms we're modeling.  And the 
mechanisms we're modeling (again, assuming this is a thread about 
modeling and not programming) are embedded in the real world.  So, to 
take this to its logical conclusion, the machines we're modeling are 
constrained solely by the ontology in which they're embedded (the laws 
of nature) and our methods and technology should, therefore, also be 
constrained solely by the ontology in which our referents are embedded.

Having said that, I am a fan of method in one sense.  When you 
communicate, you should be as explicit as possible in the service of 
getting your point across.  I'm a vague person who usually makes vague 
statements.  But, that's because I am lazy (or "efficient" if you're 
generous) and only constrain my work to the extent that it should be 
constrained for the given work.  A statement of methods in, say, a paper 
should explicitly state everything you think needs to be held constant 
in order to repeat or extend your work.  Throwing in extra detail just 
to impress might help you get published; but, it's disengenuous, as is 
hooking to a method when method is unnecessary.

All my sanctimonious hoo-ha is simply saying that methods should be 
used.  And those methods should be rigorous.  But, they should most 
empatically _not_ be standard unless and until that standard method has 
been shown to cover most of the dynamics of the mechanisms we're 
modeling.  (Perhaps the 80/20 rule is adequate for clarifying "most".) 
And that means that if you're modeling living systems, the standard 
methods will come in tandem with the theory of living systems.  In the 
meantime, you're limited to small competing demes of standardization 
that naturally pop up as a discipline ages.

[grin]  So, if you're in the Repast deme, use the methods that other 
Repast deme-members use.  If you're in the Cybernetics deme, use the 
methods that the cyberneticists use.  Etc.  And be satisfied that you're 
part of a meta-heuristic algorithm that is purposefully classifying the 
space of referents that ABM's are being used to model.

That's my 3 cents.... remember that you asked for it. [grin]

--
glen e. p. ropella              =><=                Hail Eris!
H: 503-630-4505                       http://ropella.net/~gepr
M: 971-219-3846                        http://tempusdictum.com