[Comp-neuro] useful models

Ted Carnevale carnevalet at sbcglobal.net
Mon Aug 11 18:05:43 CEST 2008

Neil Burgess wrote:
 > Re: the discussion of 'realistic' and 'useful' models.
 > In practice a useful model is one that makes predictions which
 > are novel and feasible enough to convince an experimenter
 > to actually test them. This is actually quite rare

And too exclusive a definition of utility.  Two examples will suffice.

1.  As Jim Bower noted, a model can be useful by prodding
 >> one to quantify ones ignorence by indicating which of
 >> the parameters require more data.
This happens quite commonly.  In practice, such utility is often
demonstrable _before_ the model has been implemented in computable
or mathematically tractable form.

2.  Perhaps even more frequent are private insights gained by
constructing and using models, whether they be "realistically
detailed" computational or simplified mathematical abstractions.
The design of hypotheses and experiments to test them depend
critically on the judgment and insight of the investigator.
However, in a highly interdisciplinary field like neuroscience,
many highly productive investigators have serious gaps of
knowledge and understanding that can be revealed and remedied by
modeling.  That said, ego being what it is, don't expect to hear
public confessions of private epiphanies.


More information about the Comp-neuro mailing list