[Vp-integration-subgroup] [Vp-reproduce-subgroup] "Models are not consistently licensed"

[William Waites]

> For example, if the model is calibrated against data that
> is licensed with a CC-BY-SA-ND-NC license, the model cannot
> be published under CC0.(?)

Interesting. This seems to hinge on the “ND” part. Are 
model parameters derivative works of data in the sense
of copyright law? I doubt it, in the same way that if I
write a book review or a summary it is not a derivative
work of the book. So I don’t think this is the case. I
personally would have no compunction about it. But
this strays into armchair lawyering and there might be
subtleties.

Another example. I worked on a model that was calibrated
against individually identifiable data that cannot be
released at all. It is perfectly well allowed to publish
the posterior distributions for parameters even though
the data that gives rise to them is unavoidably
proprietary.

In any event, the model can always be published under
whatever license. It is very important to distinguish
the model (software) from the parameters (data). This is
partly a question of nomenclature, but thinking of
parameters as part of the model is a confusion.

Cheers,
-w