[open-science] "open science" definition?

Mon Oct 13 15:26:30 UTC 2014

Tom Roche Sun, 12 Oct 2014 23:32:15 -0400 [1]
>> I assert that we can, to a first approximation, model a scientific study like a classic *x pipeline[2]:

>> * input > transform > output

>> * open output (data or analysis) from any one study may become an input for any number of subsequent studies

>> [For] characterization of openness, inputs and outputs are equivalently data. Unless I'm missing something, the Open Definition (or similar) should apply to both--no?

Peter Murray-Rust Mon, 13 Oct 2014 06:54:50 +0100 (rearranged) [3]
> [Jean-Claude Bradley and others including me] also were worried that "Open Access" had become so vague, variable and [diluted] that it could mean whatever you wanted. The same c/would happen for science.

Jenny Molloy has already stated the problem with that argument:

Jenny Molloy Sun, 12 Oct 2014 19:57:06 +0100 [4]
>>> multiplicity of meaning and lack of clear definitions increase the likelihood that the term is misused and it's harder to call this out. Open Access was defined by the Budapest Open Access Initiative but the term is frequently applied to terms of access that are not BOAI-compliant.

PMR's position reminds me of arguments that were made about open-source software licensing 30 years ago: software is too diverse in creation and application, licenses will multiply. Which they have, but we have a better world (and, for coders, profession) because there *are* defined licenses: without them, FLOSS would be completely meaningless, and the social movements that have coalesced around that (defined) concept probably would have not.

To distill the objection: how will *not* providing one or more definitions *reduce* meme mutation? This will *certainly* happen in the absence of framing, such as the OKF could usefully (IMHO) provide. If I perform a study in accordance with a definition of "open science" and someone claims that term is meaningless, I point to the definition and (to paraphrase Johnson) "refute it thus."

>> That leaves characterization of the openness of a study's transform(s). To a first approximation, we can separately characterize every transform as non- or computational. For computational science, I'm assuming we could leverage prior art on openness of

>> * source codes and their repositories

>> * source platforms (e.g., the hardware, OS, or other software required to run the sources)

>> I know much less about openness of non-computational protocols (I'm just a coder who works on environmental models) but assume (absent contradiction from someone who actually *knows* about this space :-) that their openness has been defined by one or more domain experts

> we decided then that "Open Science" was too broad a term to be useful in guiding practice. For many people the laboratory aspects of science and its communication are sufficiently different that one term didn't cover them.

I see no problem with refinements of the above model, whether as additional factors become modelable (e.g., social aspects, which could be accomodated by versioning the model) and to adapt the model to specific domains (e.g., for domain-specific badging such as Open Drug Discovery, noted in PMR's post[3]. And I continue to claim that "transformation of inputs to outputs" characterizes (at a high level) an aspect of scientific studies significant not for characterization of their openness.

> J-C coined the term "Open Notebook Science" [ONS] for the practice of science with immediate visibility to the whole world. The term is precise ("no insider knowledge"),

s/precise/useful/ . To the extent "no insider knowledge" is precise, it's false, in the same sense that no model is completely faithful to the domain it targets. (The point of modeling is computational tractability, for either hardware or wetware, and a publication of *everything* that happened during the course of a study is probably impossible, and certainly not worth consuming. Framing is inevitable, necessary--and problematic, which is why we seek to provide guidance.) That being said, I agree that "no insider knowledge" is a valuable starting point for a definition of an open transform, and suspect that other aspects of ONS (as summarized in slide 40 of PMR's presentation "Open Data and Open Notebook Science"[5], and following slides) would be useful.

FWIW, Tom Roche <Tom_Roche at pobox.com>

[1] https://lists.okfn.org/pipermail/open-science/20141013/003552.html
[2] https://en.wikipedia.org/wiki/Pipeline_%28Unix%29
[3] https://lists.okfn.org/pipermail/open-science/20141013/003553.html
[4] https://lists.okfn.org/pipermail/open-science/20141012/003550.html
[5] http://www.slideshare.net/petermurrayrust/osbrazil