The scientific paper – the actual form of it – is to this day used by researchers to communicate results, despite being essentially the same as the paper version created 400 years ago, and that is limiting in a world where communication has evolved into a plethora of digital channels in all shapes and formats. The “notebook” is an interactive, digital replacement that could circumvent essentially all limitations for a piece of paper (or a pdf file). That is what is argued in a piece in The Atlantic.
As a scientists I think not only that but many practices and tools are obsolete; I don’t think however that one standard of communicating science is needed to replace the paper any more than a standard for any other aspect of it – the claim that there is a clear option to replace the scientific paper format is pretty weak. In fact, the piece is less about the obsolescence of the paper and more about the virtues of the notebook; it also conflates some of its issues with the crisis of reproducibility of science – most biological and medical research is still analyzed with correlations, linear regression and ANOVA, (or worst, statisticsless plots – some are even bars!) and people are lucky if the researchers understand the methods they are using. Sloppily written software, another culprit called out in the piece, is of course worse than solid software, but people who understand statistical methods enough to write their own software are probably more part of solution than of the problem, at least in what concerns the analyses.
There is no arguing that the paper has all sorts of flaws: it is not interactive, it rarely describes all the details necessary to reproduce all the results, it has limited space, it is often dry and boring (even when in concise form and without the raw data tables), it is flat. But today the paper does not exist to fulfill all of these roles, it is more of a summary of years of research – to “tell a story”, as many people put it – and all scientists know that really learning something takes much more than reading a methods section. I heard a prominent pioneer of open access say we need to stop the myth that people read papers; I agree, most exist only to show the results, many are pretty bad at it anyway, and then they have the other sections because they have to. That is further amplified in glamour journals, which often read more like sensationalist tabloids descriptions of the actual research process. Is the scientific paper fraudulent?, Peter Medawar asked over 50 years ago; and already then he answered in the positive. That doesn’t mean the paper is useless either, but it is probably more of a by-product of all the things that are wrong in science (like in the world), than the cause of the trouble.
The notebook does not solve all those issues either, even if the work is mostly computational. It was also suggested more than once that scientific communication should be more like GitHub, but anyone who takes that at face value as a solution probably understands neither GitHub nor scientific publishing. More comprehensive ways of describing scientific work are always welcome (that is why I spend time on my own page, so I can describe what cannot be put into the methods, and give nitty gritty details of things I may have struggled with myself – which are formally in the paper and references therein, but require a lot of work to go through and actually get things going). I have used notebooks extensively at one time or another, from Mathematica and Maple, to IPython/Jupyter, through Sage and Maxima, and I think they can be very useful for interactive work as well as presentation purposes, but not necessarily for standalone communication any more than some well-commented Python code could be, for me at least. Maybe it is other scientists’ preferred method of writing or reading research, which is fine, but it might as well be worse than the paper in some cases, so why try to replace it entirely?
The Atlantic piece is correct in all the things the paper is not, and in some things the notebook is, but as the frontier of human knowledge science is extremely complex, and the form of the communication of its results is only one of many fronts (albeit an important one) where problems arise. Within there are issues of access to the publications and results, reproducibility based on published methods (which level of detail and thoroughness of the description should be there?), and sustainability of the publication system. Beyond that there are other large fronts like science funding, and the actual research activity being probably the main one. Many of these are probably static for way more than 400 years; the paper is not the only thing that needs urgent rethinking, although it just may as well be a good place to start.
(a version of this post appears on Medium, you can leave comments there or reach me on twitter)
-- caetano,
April 9, 2018