Imperial College London

Henry Rzepa

For perhaps ten years now, the future of scientific publishing has been
hotly debated. The traditional models are often thought to be badly
broken, although convergence to a consensus of what a better model
should be is not apparently close. But to my mind, much of this debate
seems to miss one important point, how to publish data.

Thus, at one extreme is [COAlition
S](https://www.scienceeurope.org/coalition-s/), a model which promotes
the key principle that "*after 1 January 2020 scientific publications on
the results from research funded by public grants provided by national
and European research councils and funding bodies, must be published in
compliant Open Access Journals or on compliant Open Access Platforms*."
This includes ten principles, one of which "*The 'hybrid' model of
publishing is not compliant with the above principles*" has revealed
some strong dissent, as seen at
[forbetterscience.com/2018/09/11/response-to-plan-s-from-academic-researchers-unethical-too-risky](https://forbetterscience.com/2018/09/11/response-to-plan-s-from-academic-researchers-unethical-too-risky/) I
should explain that hybrid journals are those where the business model
includes both institutional closed-access to the journal *via* a
subscription charge paid by the library, coupled with the option for
individual authors to purchase an Open Access release of an article so
that it sits outside the subscription. The dissenters argue that non-OA
and hybrid journals include many traditional ones, which especially in
chemistry are regarded as those with the best impact factors and very
much as the journals to publish in to maximise both the readership,
hence the impact of the research and thus researcher's career prospects.
Thus many (not all) of the American Chemical Society (ACS) and Royal
Society of Chemistry (RSC) journals currently fall into this category,
as well as commercial publishers of journals such as *Nature*, *Nature
Chemistry,Science,* *Angew. Chemie*, etc. 

So the debate is whether funded top ranking research in chemistry should
in future always appear in non-hybrid OA journals (where the cost of
publication is borne by article processing charges, or APCs) or in
traditional subscription journals where the costs are borne by those
institutions that can afford the subscription charges, but of course
also limit the access.  A measure of how important and topical the
debate is that there is even now a [movie devoted to the
topic](https://paywallthemovie.com) which makes the point of how
profitable commercial scientific publishing now is and hence how much
resource is being diverted into these profit margins at the expense of
funding basic science.

None of these debates however really takes a close look at the nature of
the modern research paper. In chemistry at least, the evolution of such
articles in the last 20 years (\~ corresponding to the online era) has
meant that whilst the size of the average article has remained static at
around 10 "pages" (in quotes because of course the "page" is one of
those legacy concepts related to print), another much newer component
known as "**Supporting information**" or
SI[^♥^]{style="color: #ff0000;"} has ballooned to absurd sizes. It can
reach 1000 pages\[cite\]10.1021/jacs.6b13229\[/cite\] and there are
rumours of even larger SIs. The content of SI is of course mostly data.
The size is often because the data is present in visual form (think
spectra). As visual information, it is not easily "inter-operable" or
"accessible". Nor is it "findable" until commercial abstracting agencies
chose to index it. Searches of such indexed data are most certainly
"closed" (again depending on institutional purchases of access) and not
"open access". You may recognise these attributes as those of **FAIR**
(Findable, accessible, inter-operable and re-usable). So even if an
article in chemistry is published in pure OA form, in order to get FAIR
access to the data associated with the article, you will probably have
to go to a non-OA resource run by a commercial organisation for profit.
Thus a 10 page article might itself be OA, but the full potential of its
1000+ page data (an elephant if ever there was one) ends up being very
much not OA.

You might argue that the 1000+ pages of data does not require the
services of an abstracting agency to be useful. Surely a human can get
all the information they want from inspecting a visual spectrum? Here I
raise the future prospects of AI (artificial intelligence). The \~1000
page SI I noted above\[cite\]10.1021/jacs.6b13229\[/cite\] includes
*e.g* NMR spectra for around 70 compounds (I tried to count them all
visually, but could not be certain I found them all). A machine, trained
to identify spectra from associated metadata (a feature of FAIR), could
extract vastly more information than a human could from FAIR raw data^‡^
(a spectrum is already processed data, with implied information/data
loss) in a given time. And for many articles, not just one. Thus FAIR
data is very much targeted not only at humans but at the AI-trained
machines of the future.

So I again repeat my assertion that focussing on whether an article is
OA or not and whether publishing in hybrid journals is to be allowed or
not by funders is missing that 100-fold bigger elephant in the room. For
me, a publishing model that is fit for the future should include as a
top priority a declaration of whether the data associated with it is
FAIR. Thus in the [Plan-S ten
principles](https://www.scienceeurope.org/wp-content/uploads/2018/09/Plan_S.pdf),
FAIR is not mentioned at all. Only when FAIR-enabled data becomes part
of the debates can we truly say that the article and its data are on its
way to being properly open access.

------------------------------------------------------------------------

^‡^The FAIR concept did not originally differentiate between processed
data (*i.e.* spectra) and the underlying primary or raw data on which
the processed data is based. Our own implementation of FAIR data
includes both types of data; raw for machine reprocessing if required,
and processed data for human interpretation. Along with a rich set of
metadata, itself often created using carefully designed workflows
conducted by machines.

[^♥^]{style="color: #ff0000;"}The proportion of articles relating to
chemistry which do not include some form of SI is probably low. These
would include articles which simply provide a new model or
interpretation of previously published data, reporting no new data of
their own. A famous historical example is Michael Dewar's
re-interpretation of the structure of stipitatic
acid\[cite\]10.1038/155050b0\[/cite\] which founded the new area of
non-benzenoid aromaticity.

::: {.wp_orcid_field}
[https://orcid.org/0000-0002-8635-8390](https://orcid.org/0000-0002-8635-8390){target="_blank"
rel="author"}
:::


For perhaps ten years now, the future of scientific publishing has been hotly debated. The traditional models are often thought to be badly broken, although convergence to a consensus of what a better model should be is not apparently close. But to my mind, much of this debate seems to miss one important point, how to publish data.

Publicaciones de Rogue Scholar

Open Access journal publishing debates – the elephant in the room?