Re: slightly off-topic: support open source for publically-funded research

[Marcus G. Daniels]

>>>>> "GR" == glen e p ropella <address@hidden> writes:

OI> The longer-term benefit is that the software can be studied and
OI> reviewed in the same way as the other parts of scientific
OI> research. Software flaws can cause as misleading results in the same
OI> way as sloppy protocols or faulty math. Exposing all the scientific
OI> process to peer review can only lead to better science.

GR> Besides, the petition is specious because open source is not
GR> "...the software equivalent of peer-reviewed publication of
GR> research results". 

GR> Open source is the equivalent of standardized tools in scientific
GR> research.  The software, itself, is not the research.  The software is
GR> one (and only one) description of one of the pieces of experimental
GR> equipment used.

In a peer-reviewed scientific publication, I don't believe agent-based
models of any complexity will be described in sufficient detail to
reliably reproduce results.  Often there will be too much detail and
funny corner cases, and thus it is important to have access to the
source code of the agent-based model.  In these complex cases, the
implementation of the model is not just an instrument but a crucial,
bottom-line representation of the model.  Much the flexibility of
agent-based modeling comes from the fact you can look at the dynamics
of mechanisms that are otherwise hard to represent using conventional
mathematical tools.  This flexibility comes at the cost of clarity,
but it may be worth it.

In cases where the models are simple enough to fully describe, then
the instruments are not a big deal (in terms of work) and can and
should be reimplemented just for the sake of clarity and
reproducability.

[Unfortunately, my experience is that a lot of people doing agent-based
modeling are not at all comfortable (or willing) to read code, and
thus little insight is gained by having access to it.]

GR> But requiring that the source code be provided is like requiring that
GR> a very detailed specification of a particular piece of hardware be
GR> required for another type of scientific experiment.  Rather than
GR> saying "We used the Tektronix XYZ model oscilloscope" one would have
GR> to provide a detailed description of the particular oscilloscope used,
GR> which is a pretty ridiculous requirement.

Well, an oscilloscope is a well understood thing.  It is less clear,
say, that a specification of concurrent scheduling for different agent
types is a language that is sufficiently common and concise to
substitute as an instrument in the same way an oscilloscope does.  
I hope that will happen eventually..

                  ==================================
   Swarm-Support is for discussion of the technical details of the day
   to day usage of Swarm.  For list administration needs (esp.
   [un]subscribing), please send a message to <address@hidden>
   with "help" in the body of the message.